You are on page 1of 26

3 Sea shells, ancient beads and Middle

Stone Age symbols

3.1 Blombos Cave: a veritable treasure trove


From an archaeological perspective, Blombos Cave rates as a veritable treasure
trove. The archaeologist Steven Mithen, for instance, clearly impressed by
the artefacts that have been found at the cave, states that Blombos Cave is
‘the most important currently known archaeological site for understanding
the origin of modern thought and behaviour – and, by implication, language’
(2005: 250). Among the finds made by Christopher Henshilwood and his
colleagues there are engraved ochres, bifacial points, bone tools and, notably,
shell beads (Henshilwood and Dubreuil 2009: 49–51, 2011: 372–4). These
beads, they claim, were worn by the people who inhabited Blombos Cave in
the Middle Stone Age. And, from data about properties of these prehistoric
shell beads they, as we noted in Chapter 1, infer the following:
Conclusion about language of Blombos inhabitants
The humans who inhabited Blombos Cave some 75,000 years ago
had fully syntactic or modern language (d’Errico et al. 2005: 19–20;
Henshilwood and Dubreuil 2009: 45; Henshilwood et al. 2004: 404).
The inference with this conclusion – henceforth the ‘Blombos inference’ –
deserves close analysis for various reasons; we will now consider three of these.
First, if sound, the inference would bear in an interesting way on the
question of when modern language emerged. That is, it would push the
emergence of modern language back by roughly 25,000–30,000 years. And,
in so doing, it would run counter to the belief – held by the archaeologist
Richard Klein (2000: 27), the palaeoanthropologist Ian Tattersall (1998a:
230–1, 1998b: 23–5) and others – that modern language emerged as late as
50,000 to 40,000 years ago.
Second, the Blombos inference is representative of a major category of
inferences drawn by archaeologists about the evolutionary emergence of lan-
guage. Starting out from data about properties of artefacts, and moving on
stepwise from there, these inferences typically end up by ascribing some stage
of language to some group of prehistoric humans. These inferences owe their

29

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
30 Language Evolution

potential archaeological significance to two fundamental assumptions: (i) that


the key criterion for modern human behaviour is the use of symbolism to
organise that behaviour (Henshilwood and Marean 2003; Wadley 2001), and
(ii) that the stage of language needed to organise behaviour symbolically is the
one referred to as ‘fully syntactic language’ or ‘modern language’ (Henshilwood
and Marean 2003; McBrearty and Brooks 2000; Mellars 1998a, 1998b). In these
terms, then, fully syntactic language is held to be distinctive of modern human
behaviour, also referred to as ‘human modernity’ (Henshilwood and Dubreuil
2009: 65). Hence, if sound, the conclusion ascribing fully syntactic language
to the humans who inhabited Blombos Cave some 75,000 years ago would
imply that they were behaviourally modern people. The Blombos inference,
what is more, is not just like any other member of the category of archaeological
inferences at issue; it is arguably the most fully articulated one. This means that
the Blombos inference is important also for what it can reveal about the
properties of this category of inferences. For example, if the Blombos inference
were found to be less than sound for non-accidental reasons, this would reflect
on ways in which its whole category of archaeological inferences is problematic.
Third, the soundness of the Blombos inference bears on the tenability of an
intriguing view of the importance of archaeological material as a source of
information about the origin of language. In Francesco d’Errico’s own words,
this is the view that archaeologists and palaeoanthropologists ‘probably retain
in their hands the best, not to say the only, information’ to answer the question
of ‘what syntactical language is and where it arose’ (Számadó et al. 2009:
231).1 Interestingly, d’Errico has co-authored with Henshilwood various art-
icles in which the Blombos inference is favourably presented (e.g., d’Errico
et al. 2005; Henshilwood et al. 2004). d’Errico’s view of the importance of
archaeological information for work on language origins would sustain clear
damage, therefore, if the Blombos inference failed to bear it out.
For various reasons, then, the Blombos inference warrants close analysis.
But what would such an analysis involve? Precisely what the analysis of any
window inference involves. Such an analysis is guided by the following
questions: (i) ‘How is the inference structured?’ (ii) ‘How sound is the
inference?’ (iii) ‘If the inference is less than sound, why is that so?’ (iv) ‘What
can be done to overcome the problems with the inference?’ In such an analysis
of the Blombos inference, it will be useful to start out from a schematic
representation of the three steps in terms of which the inference is made. For
just this, see Figure 3.1.
I use the term the ‘shell-bead window’ to refer to the window that allows
one to take the inferential steps represented by the arrows B, D and F in
Figure 3.1. These three inferential steps are discussed in Sections 3.2–3.4,
respectively.2 Before we turn to the specifics of the inferential steps repre-
sented by B, D and F, notice that the shell-bead window is a correlate window,

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 31

Conclusion/
Conclusion/ Conclusion
Data assumptions
assumptions that Blombos
about about
about beads Cave
properties of Inferential Inferential symbolic Inferential
worn by inhabitants
Middle behaviour of
step Blombos step step had fully
Stone Age Blombos
Cave syntactic
tick shells Cave
inhabitants language
inhabitants
A B C D E F G

Figure 3.1 Structure of a composite inference about the language of


Blombos Cave inhabitants.

in the sense set out previously under Part II. That is, the tick shells represented
in block A of Figure 3.1 are taken to be correlates of the beads represented in
block C; these beads are taken to be correlates of the symbolic behaviour
represented in block E; and this behaviour is taken to be a correlate of fully
syntactic language represented in block G.

3.2 From shells to beads

3.2.1 The grounds


Let us consider then the first inferential step of the Blombos inference – shown
schematically as arrow B in Figure 3.1. In terms of this step, it is inferred from
data about certain properties of a number of Middle Stone Age tick shells that
these shells were worn as beads by the humans who inhabited Blombos Cave
some 75,000 years ago. These data are about forty-one shells of a species of
sea snails; technically, this is the scavenging gastropod known as Nassarius
kraussianus. As detailed in the literature (d’Errico et al. 2005; Henshilwood
and Dubreuil 2011: 372–5; Henshilwood et al. 2004), these data include the
following:

Data about MSA tick shells


(a) The shells are approximately 75,000 years old.
(b) As for their physical properties: (i) the shells are perforated; (ii) the type
of perforation in the shells is rare in nature and difficult to explain as the
result of natural processes; (iii) the shells have flattened facets; and (iv)
four of the shells show microscopic traces of red ochre on their insides
and surfaces.
(c) As for their location: the shells were found kilometres away from the
estuaries in which the molluscs were likely to have occurred.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
32 Language Evolution

(d) Thirty-three shells were found in six groups of two to twelve beads,
with all six groups being recovered in a single excavation day, and all
six coming either from a single square or from two adjacent sub-squares
of the cave.
(e) Shells found in the same group are similar in regard to their adult size,
their shade and their type of perforation.
The properties of the shells are believed to be interlinked in a way that
Henshilwood and Dubreuil (2011) describe speculatively as follows in a recent
overview of the material culture of Homo sapiens in southern Africa:
An analysis of 41 of the recovered Nassarius kraussianus ‘tick’ shells from Blombos
Cave shows that they were carefully pierced with a bone tool to create a keyhole
perforation (d’Errico et al. 2005; Henshilwood et al. 2004). These perforations are
anthropogenic and deliberate. The beads [sic] were then strung, perhaps on cord or
sinew, and then worn as a personal ornament. . . . Repeated rubbing of the beads against
one another and against the cord resulted in discrete use-wear facets on each bead that
are not observed on these shells in their natural environment . . . Microscopic residues of
ochre occur inside some of the beads and may result from deliberate coloring or from
transfer when worn (d’Errico et al. 2005; Henshilwood et al. 2004). (Henshilwood and
Dubreuil 2011: 374)

The question, of course, is whether the data about the shells form an adequate
basis for the inferential step at issue. This question springs from a fundamental
soundness condition that inferences about language evolution need to meet
in order to be sound. It is formulated in Section 2.5 as the Groundedness
Condition, which says that an inferential step leading to a conclusion about
language evolution needs to be grounded in firm data or empirical assumptions
about properties of phenomena that are well understood.
With reference to Figure 3.1, the question is then: ‘How well is the inferen-
tial step shown as arrow B grounded in the data shown in box A?’ In this
regard, some sceptical questions have been raised about the origin of the holes
in the shells (Holden 2004). Yet, from the publications involved (e.g., d’Errico
et al. 2005; Henshilwood et al. 2004), it is clear that the data forming the basis
of inferential step B have been collected, analysed, checked and recorded with
conspicuous care. This includes the use of tests and experiments for establish-
ing the correctness of some observations. For instance, techniques based on
thermoluminescence and optically stimulated luminescence were applied to
single aliquots and 4800 individual quartz grains from the MSA layer in which
the shells were recovered. This was done to accurately date this layer and, by
implication, the shells found in it. In addition, Henshilwood and his colleagues
tested their observations about the location, morphology and size of the holes
in the shells by experimentally piercing the wall of a number of shells with
three types of tools with which the inhabitants of Blombos Cave could have

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 33

made the perforations.3 The research group, in short, has taken great pains to
make sure that the data about their finds from the cave are correct. This view is
in line with Iain Davidson’s (2011: 383) observation that: ‘The finds from
Blombos Cave have been meticulously excavated and documented.’

3.2.2 The warrant


Like all inferences drawn in empirical work about the evolution of language,
the inferential step shown as B in Figure 3.1 needs to meet a second soundness
condition. It was stated in Section 2.4 above as the Warrantedness Condition.
And it says in effect that an inferential step leading from data or assumptions
about a phenomenon of one kind to a conclusion about a phenomenon of a
different kind needs a licence to move across the ontological gap between
these phenomena. The question about step B is: ‘Why is it warranted to infer
from the data about MSA tick shells listed in (a)–(e) above that they were worn
as beads?’ Shells and beads are clearly different kinds of objects belonging to
different ontological domains – shells to that of material objects, and beads
to that of cultural objects manufactured by humans for some purpose. The
distinction between these two domains is obscured by the compound expres-
sion ‘shell beads’ used by Henshilwood, d’Errico and other archaeologists.
The shells excavated by them are physical objects; the claim that those shells
were worn as beads, that is, that they served as cultural objects, must of
necessity be an inference. Henshilwood and d’Errico are correct, therefore,
to point out that evidence is required for ‘[d]etermining whether N. kraussia-
nus shells . . . are beads’ (d’Errico et al. 2005: 10). What the Warrantedness
Condition lays down is that such evidence must be of a relevant kind.
From the discussion in Section 2.4, it is clear that the warrant needed for the
inferential move from MSA sea shells to beads cannot be the mere stipulation
that data about such shells provide evidence for such beads. The inferential
step shown as B in Figure 3.1 needs, instead, a warrant underpinned by a
bridge theory; a bridge theory which gives an account of how properties of tick
shells are interrelated with properties of beads or beadworks. Henshilwood and
d’Errico do not give the assumptions of the required bridge theory in an
explicit form. From their discussion (d’Errico et al. 2005), however, it is
possible to gauge some of the assumptions which they had to make in order
to move inferentially from the domain of tick shells to that of MSA beads.
When we abstract away from less relevant details, these bridge assumptions
are seen to include the following:

The step from MSA shells to beads: bridge assumptions


(a) MSA shells are probably beads if: (i) the shells were collected, trans-
ported and accumulated by humans; (ii) the shells have perforations that

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
34 Language Evolution

were made by humans; (iii) the shells have properties that were caused
by use-wear (e.g. by rubbing on something); (iv) the shells exhibit
traces of ochre resulting from tool use by humans.
(b) A group of MSA shells are probably beads forming part of the same
beadwork if: (i) the shells are similar in regard to size, shade, type of
perforation and other physical properties; (ii) the shells are recovered in
the same excavation period in the same square or adjacent sub-squares
of the site.4
In the article referred to above, Henshilwood and Dubreuil (2011: 374) single
(a)(iii) out as the core assumption: ‘The use-wear patterns are the principal
factor that defines the shells as beads.’
The bridge assumptions stated in (a) and (b), then, enable Henshilwood and
d’Errico to infer from their data about the forty-one tick shells listed in (a)–(e)
above that these shells were beads that formed part of a larger beadwork. The
soundness of the inference depends on the strength of the support that these
bridge assumptions individually and collectively have. It is not clear from
Henshilwood and d’Errico’s discussion how this support comes about or how
strong it is. Since these bridge assumptions are being invoked in empirical
work, they ought to be non-ad hoc and cannot be mere stipulations. Nor is it
clear what significance should be attached in this regard to the non-specific
reference made by Henshilwood and d’Errico (Henshilwood et al. 2004: 404,
n. 7) to a work on the history of beads by Dubin (1987). The Blombos inference
will be shown below to have been criticised on various grounds; none of the
criticisms, however, has been aimed at the inferential step from shells to beads.
All in all, then, the assumptions stated as (a) and (b) above give the right kind of
warrant for the conclusion that the shells in question are beads.

3.3 From beads to symbols


This brings us to the second inferential step of the Blombos inference –
represented in Figure 3.1 by arrow D. In terms of this step, it is inferred from
assumptions about the beads (or beadworks) worn by the MSA inhabitants of
Blombos Cave that the beads were symbols or, equivalently, that the wearing
of the beads was a form of symbolic behaviour. This inferential step is needed
since, as observed by Lyn Wadley (2001: 270) amongst others, artefacts such
as beads are not automatically imbued with symbolism. This means that from
the fact – should it be one – that the beads were worn by MSA humans it does
not necessarily follow that these beads were symbols. The point has weighty
consequences to which I will return below.
In the case of the inferential step from shells to beads, we have seen,
Henshilwood and d’Errico answer in considerable detail the question: ‘What

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 35

data about tick shells does that step start from?’ In the present case,
the question is: ‘What are the data and/or empirical assumptions about such
beads from which the authors infer that these beads were symbols?’ Or, in
other words: ‘What is the content of box C in Figure 3.1 in which the
inferential step at issue is grounded?’ In early publications (d’Errico et al.
2005; Henshilwood et al. 2004), the authors do not explicitly present the
required data or assumptions as such. What one does find there is a number
of tangential remarks on the great difficulty in determining whether MSA
artefacts were symbols and, more generally, on the criteria for symbolhood.
As for determining ‘how symbolism is reliably recognized in an artefact’ –
to use a formulation of Henshilwood et al. (2004: 404) – one finds remarks
such as the following:
A primary problem when examining evidence for ‘modern’ behaviour is the criteria
used for assessing markers interpreted as ‘symbolic’: In particular, how is the distinction
made between potentially symbolic objects deliberately crafted by humans from those
produced by functional activities or natural processes? (d’Errico et al. 2003: 17–18)5
From these remarks it is clear that Henshilwood and d’Errico see the problem of
determining the symbolhood of an object as non-trivial. They do not, however,
explicitly present the conceptual or methodological means which in their view
should be used for making the determination. What they do, instead, is to
address the problem obliquely by indicating what they take symbols to be. That
is, referring to work of other scholars, they take their stand as follows:
A key characteristic of all symbols is that their meaning is assigned by arbitrary,
socially constructed conventions (Chase & Dibble 1987). Perhaps the greatest benefit
of symbolically mediated behaviour is that it permits the storage and display of infor-
mation external to the human brain (Donald, 1991; Wadley, 2001; Henshilwood &
Marean, 2003; Hovers et al., 2003). (d’Errico et al. 2005: 4)
Symbols are representative of social conventions, tacit agreements, or explicit codes
that link one thing to another and are mediated by some formal or merely agreed-upon
link irrespective of any physical characteristics of either sign or object (Deacon
1997: 70). (Henshilwood and Marean 2003: 635)
In Henshilwood and d’Errico’s view, then, symbols have three properties:
(i) symbols are objects that have a meaning; (ii) the meaning of symbols is
assigned by arbitrary social conventions, tacit agreements or explicit codes;
and (iii) symbols make possible the storage and display of information external
to the human brain. In terms of this view, to be able to infer that the Blombos
beads had the status of symbols, one would need to present specifics of – at
least – three factors: (i) the meaning(s) that these beads had for the inhabitants
of the cave; (ii) the social conventions etc. by which these beads were assigned
the meanings at issue; and (iii) the information that was stored in or displayed
by these beads in the event that this information differed from these meanings.6

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
36 Language Evolution

In the absence of such specifics, a question arises about whether the


inferential step from beads to symbols meets the Groundedness Condition. In
Henshilwood and d’Errico’s own judgement, interestingly, the amount of data
that can be drawn on in efforts to reconstruct whatever symbolic meaning held
good in the southern Africa MSA is strictly limited. Thus, they maintain that
[g]iven the very limited regional data available in southern Africa and the problems
with unravelling and interpreting symbolic meaning in such deep time, this remains an
unrealistic goal [i.e., the goal of verifying for the southern Africa MSA whether putative
symbolic communities intentionally communicated cultural identities]. (d’Errico et al.
2001: 317)
As shown above, Henshilwood and d’Errico have taken great care in collect-
ing, analysing and checking the data needed for grounding the inferential step
from shells to beads. It is striking therefore that, in making the inferential
step from beads to symbols, they have not been similarly careful about the data
about the beads that are needed for grounding this step. A possible explanation
for this might be that, while adopting the view that it is not the nature of
artefacts in general to be symbols, Henshilwood and d’Errico make the
following assumptions:
Assumptions linking beads, ornaments and symbols
Assumption 1: the beads were worn as ornaments by the MSA
inhabitants of Blombos Cave.
Assumption 2: ‘personal ornaments and art are unquestioned expres-
sions of symbolism that equate with modern human behaviour’
(d’Errico et al. 2005: 4).
The inferential step from beads to symbols would then be grounded in
Assumption 1. And the step itself would be warranted by Assumption 2, which
is believed to serve as a bridge between the domain of ornaments and that of
symbols.
On a first analysis, then, it might seem as if the inferential step from the
Blombos beads to symbols met both the Groundedness Condition and the
Warrantedness Condition. This would be the case in fact, however, only if
appropriate justification were furnished for Assumptions 1 and 2. In particular,
it would be unacceptable for these assumptions to be just dogmatic stipulations
or unsupported speculations. So the status of these two assumptions is what we
need to look into next.
According to Lyn Wadley (2001: 2003) ‘probably all archaeologists’ are
agreed on Assumption 2. Though this agreement may be no more than the
expression of a deep-seated disciplinary belief, it does provide some sort of
support for Assumption 2. Challenged by various archaeologists, Assumption 1,
however, has been highly contentious. Some of these challenges have come in a
discussion of Henshilwood and Dubreuil’s (2011) article on the Still Bay and

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 37

Howiesons Poort. Thus, clearly not judging Assumption 1 sufficiently well


justified, Frederick Coolidge and Thomas Wynn (2011: 380) make the counter-
proposal that the beads ‘may have been tokens on a tallying device, a string of
beads to keep track of quantities’. They (2011: 380) disagree with the link that
Henshilwood and Dubreuil assume to exist between ‘ornament and conscious
symbol use’, maintaining that the beads ‘have no explicit or implicit meaning
beyond a simple association of appearance and social life’. In Coolidge and
Wynn’s view, the beads definitely were not symbols but could instead have
been ‘indexes of membership of some vague group’.
In commenting on Henshilwood and Dubreuil’s article, Malafouris (2011)
questions the basis on which symbolic status has been assigned to artefacts
recovered at Blombos Cave. Thus, he maintains that Henshilwood and
Dubreuil

fall short of recognizing the problem of our limited archaeological understanding of the
nature of material signification (Malafouris 2007, 2010b). As a result, their examination
of the archaeological evidence simply reiterates the usual conventional assumptions
about the putative ‘symbolic’ character and stylistic function of certain categories of
artifacts. (Malafouris 2011: 385)

And he (2011: 385) goes on to ask rhetorically ‘why, for example, does a
series of deliberately incised lines [in a piece of ochre] suggest or embody
symbolic meaning? When and how are the markings symbolic?’ On his
judgement, Henshilwood and Dubreuil have not given a satisfactory answer
to these basic questions.
How, then, does Malafouris himself view the significance of the Blombos
beads? He ‘cannot see anything ‘parsimonious’ about the inference that beads
can act as symbols’ (2011: 386). Nor does he recognise the beads ‘as reflecting
an already symbolically equipped human mind’; he views them, instead, as
having been ‘potentially powerful, enactive material signs (Malafouris 2007,
2008) capable of bringing forth a new conception of selfhood and perspective
taking’ (Malafouris 2011: 386).
Henshilwood and Dubreuil (2011: 390ff.) have responded to some of the
criticisms and counterproposals made by Coolidge and Wynn, Malafouris and
the other scholars who have commented on their article. They (2011: 391)
contend, for instance, that ‘the interpretation of shells as personal ornaments
seems more reasonable, because the use of personal ornaments is ubiquitous
among human foragers, while that of tallying devices is not’. The strength of
this contention – and other points argued by Henshilwood and Dubreuil – will
become clear in a debate that is bound to continue. At this stage, the inferential
step from beads to symbols – marked in Figure 3.1 as C – remains controver-
sial, and so of dubious soundness. To remove these doubts, the grounding and
the warrant of the step from beads to symbols require better underpinning.7

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
38 Language Evolution

3.4 From symbols to fully syntactic language


Represented by arrow F in Figure 3.1, the third inferential step making up the
Blombos inference is that from symbols or symbolic behaviour to language.
More specifically, from assumptions about the symbolic behaviour of the MSA
inhabitants of Blombos Cave, it is inferred that they had modern or fully
syntactic language. From the viewpoint of the evolution of language, this step
is particularly interesting. So the question is: ‘How sound is this third inferen-
tial step?’ To be able to arrive at an answer to this question, we have to
ascertain whether this step is well grounded, whether it is appropriately
warranted, and whether the conclusion yielded by it is sufficiently pertinent.

3.4.1 The grounds


What, then, are the grounds – represented as E in Figure 3.1 – for the
inferential step under consideration? Clearly, these need to be empirical data
or assumptions about the symbols or symbolic behaviour of the people who
inhabited Blombos Cave in the Middle Stone Age. Not that those data or
assumptions are explicitly presented as such in the pertinent literature. Instead,
they are referred to obliquely in remarks such as these:
Fully syntactical language is arguably an essential requisite to share and transmit the
symbolic meaning of beadworks and abstract engravings such as those from
Blombos Cave. (Henshilwood et al. 2004: 404)
Since syntactical language is the only means of communication bearing a built-in meta-
language that permits creation and transmission of other symbolic codes (Aiello, 1998),
beadwork represents a reliable proxy for the acquisition of language and fully modern
cognitive abilities by southern African populations 75,000 years ago. (d’Errico et al.
2005: 19–20)
These finds [of engravings at Blombos Cave] demonstrate that ochre use in the MSA
was not exclusively utilitarian art and, arguably, the transmission and sharing of the
meaning of the engravings relied on fully syntactical language. (Henshilwood et al.
2002: 1279)
It is widely accepted that a direct link exists between the highly symbolic nature of
modern language (i.e. its capacity to refer to past, present and future-actual or
imaginary-events) and the creation, maintenance, and transmission of the material
expression of symbols within a given human culture. (d’Errico et al. 2003: 6)
The complex syntactical language essential for encoding these symbolic referents
[provided for by some systems of external storage or representation] could have emerged
simultaneously and allowed for the negotiation of power during periods of increasing
social pressure and higher settlement densities. (Henshilwood and Marean 2003: 636)
Decoding the meaning of a design engraved on a piece of ochre or understanding why a
bone tool is crafted much more carefully than necessary for a utilitarian object is

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 39

difficult, but objects like these are strongly suggestive of the advanced levels of
symbolic thought and language that were necessary for the development of modern
behavior. (Henshilwood and Marean 2003: 636)

From these and other similar remarks, it is possible to deduce what Henshil-
wood, d’Errico and others assume the symbolic behaviour of the MSA inhabit-
ants of Blombos Cave to have involved.

Assumptions about what MSA symbolic behaviour involved


(a) Sharing the symbolic meaning (Henshilwood et al. 2002: 179; Henshil-
wood et al. 2004: 4).
(b) Transmitting the symbolic meaning (Henshilwood et al. 2002: 1279;
Henshilwood et al. 2004: 404).
(c) Creating the symbolic codes (d’Errico et al. 2005: 19–20).
(d) Transmitting the symbolic codes (d’Errico et al. 2005: 19–20).
(e) Creating the material expressions of symbols (d’Errico et al. 2003: 6).
(f) Transmitting the material expressions of symbolic codes (d’Errico et al.
2003: 6).
(g) Maintenance of the material expressions of symbols (d’Errico et al.
2003: 6).
(h) Encoding the symbolic referents (Henshilwood and Marean 2003: 636).
(i) Decoding the meaning of a design (Henshilwood and Marean 2003: 636).
The inferential step represented by F in Figure 3.1 cannot, though, be properly
grounded in these assumptions; for that, they are too vague. To ground this
step, one has to adduce specifics of how the activities listed as (a)–(i) – or at
least a subset of these – were manifested in the behaviour of the inhabitants of
Blombos Cave. On the one hand, that is, one needs to know in detail the
symbolic meanings, symbolic codes and symbolic referents; on the other hand,
one needs to know in detail how these were created, shared and transmitted.
Here, to illustrate, are four of the questions to which detailed answers are
required in the case of (a) and (b): ‘What are the symbolic meanings that
were shared and transmitted by the inhabitants of Blombos Cave?’ ‘What did
the sharing and transmitting involve?’ ‘Were these meanings shared by all
these inhabitants or only by a particular group of individuals?’ ‘By whom were
these symbolic meanings transmitted – all the inhabitants of the cave, only a
particular subgroup of them, or only individuals with a certain status?’
The details thus asked for are not to be found in the relevant writings of
Henshilwood and d’Errico. This means that, judged in terms of the Grounded-
ness Condition, the inferential step from symbolic behaviour to fully syntactic
language is less than sound.
How, then, is one to come up with a proper grounding for this inferential
step? That is, how can one obtain the required particulars of the putative

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
40 Language Evolution

symbolic behaviour of the people who inhabited Blombos Cave in the Middle
Stone Age? Like most other phenomena that might be linked to the evolution
of language, this behaviour does not speak for itself or wear its explanation on
its sleeve, as it were. Nor can it be studied by direct inspection. In empirical
work, the only means of getting to know phenomena such as prehistoric
symbolic behaviour is to form precise theories about them. Accordingly, to
get at the details required to underpin assumptions such as (a)–(i), what is
needed is a theory of the symbolic behaviour of the MSA Blombos inhabitants.
Like any theory constructed in empirical science, this theory of MSA symbolic
behaviour would have to have certain good-making properties: it would have
to be explicit, appraisable and supported by evidence. In terms of function, it
would instantiate the kind of underpinning theories labelled ‘grounding theor-
ies’ in Section 2.5.

3.4.2 The conclusion


This brings us to the third component of the inferential step from symbolic
behaviour to modern language: the conclusion represented by box G in
Figure 3.1. Like all conclusions about the evolution of language, the conclu-
sion that the inhabitants of Blombos Cave had fully syntactic language some
75,000 years ago needs to meet the soundness condition formulated as the
Pertinence Condition in Section 2.6. Recall that this condition says that a
conclusion about the evolution of language needs to be pertinent in being
about the evolution of one or more linguistic entities that are correctly identi-
fied and accurately characterised. And the question is whether the conclusion
attributing fully syntactic language to the MSA inhabitants of Blombos Cave is
sufficiently pertinent.
To determine that, certain more specific questions need to be pursued:
‘What exactly is the entity that the conclusion is about?’ ‘What are the
distinguishing properties of fully syntactic language?’ ‘How does fully syntac-
tic language differ from language that is not fully so?’ These are not questions
with mere terminological import; they are about a factual matter. For, on a
widely held view, syntax evolved gradually in terms of steps or stages. Ray
Jackendoff (2002: 238), for instance, provides for the evolution of syntax to
have started from protolanguage that is claimed to have had only rudimentary
syntax. And to have moved from there up to modern language in five partially
ordered stages.8 He lays these steps out schematically as in Figure 3.2.
The steps provided for in Jackendoff’s scenario each contributed to what he
refers to as ‘the precision and variety of expression’. Now, if it is plausible that
there were various steps or stages in the evolution of syntax, the question is
this: ‘What kind or degree of complexity would syntax have had to possess to
serve the communicative function(s) that are being attributed to the language

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 41

(Protolanguage about here)

Hierarchical phrase structure

Symbols that explicitly encode Grammatical categories


abstract semantic relations

System of inflections System of grammatical


to convey semantic functions to convey
relations semantic relations

(Modern language)

Figure 3.2 Steps in the evolution of syntax according to Jackendoff


(2002: 238). By permission of Oxford University Press.

used at Blombos Cave some 75,000 years ago?’ Alternatively: ‘Which of the
stages distinguished by gradualists such as Jackendoff would correspond to
what is referred to by the expression “fully syntactical language”?’ The answer
to this question is crucial for ruling out an interesting possibility; namely, that
the stage of syntactic evolution underlying the sentences uttered by Blombos
inhabitants some 75,000 years ago may have been one of the earlier, and hence
less complex, stages in the evolution of syntax. This question would not
dissolve if the specifics proposed by Jackendoff and others turned out to be
incorrect in regard to the number and make-up of the stages. It will continue to
be pertinent for as long as there are respectable gradualist accounts to the effect
that syntax did not emerge in its full modern complexity in one fell swoop.9
As for Henshilwood and d’Errico’s notion of ‘fully syntactical language’, they
have not offered an explicit characterisation of the linguistic entity, should it
exist, denoted by this notion. Referring to work by Wynn (1991), Henshilwood
and Marean (2003: 635) invoke an entity which they label ‘syntactical language
use’ and which they characterise as ‘a combination of grammar, semiotic ability,
and its pragmatic application’. In the relevant article by Wynn (1991), one finds
the following characterisation of the entity he speaks of as language:
Language, of course, consists of more than just grammar; indeed it is probably best to
think of language as a very complex behaviour that involves the interweaving of many

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
42 Language Evolution

components (Lieberman 1984; Chomsky 1980). These include, in addition to grammar,


a symbolic (semiotic) ability, knowledge of how to use language (Pragmatics), and the
biomechanical and neural structure of speech. (Wynn 1991: 191–2)

This characterisation of ‘language’ is problematic in several ways. Thus, it does


not distinguish between language as a cognitive entity and the use of language as
a form of behaviour (see Section 1.3.2). What is more, the reference to Chomsky
(1980) is puzzling since one of Chomsky’s most fundamental claims is that
language is not a form of behaviour. And the characterisation as such is not
informative about the nature of ‘syntactical language’ (and that perhaps was
not its purpose in the first place). An obvious paraphrase would be ‘language of
which syntax is the distinctive component’. But this paraphrase, in turn, is
uninformative in the absence of a characterisation of what syntax is. All in all,
we are left without a characterisation of the entity denoted by the expressions
‘syntactical language’ and ‘fully syntactical language’. And so, by being unclear
about what ‘fully syntactical language’ is, the conclusion ascribing it to MSA
inhabitants of Blombos Cave fails to meet the Pertinence Condition.
How, then, have Henshilwood and his colleagues been responding to this
difficulty? Rather than clarify the notion of ‘fully syntactical language’ in a
principled way, they abandoned it in later work. And, as for the defining
property of modern language, Henshilwood and Dubreuil (2009: 59) next
assigned that status to recursion.10 The ‘Rubicon of modern language’, as they
put it, ‘probably lies in recursion.’ But the seeming gain from this move of
theirs was short-lived. For, owing to reasons that will be detailed in Section
3.4.3, they were unable to show that the MSA people who lived at Blombos
Cave had a form of language that used recursion. Indeed, Henshilwood and
Dubreuil (2011: 392) seem to give up the idea that recursion is the defining
property of modern language. Going further still, they seem to abandon even
the theory of syntax that holds syntax to be the central feature of language
structure. That is, they (2011: 392) ‘are increasingly inclined to see syntax as
the outcome of a cultural evolutionary process, triggered by the evolution of
social cognition, rather than as a cognitive development of its own’. Beyond a
non-specific reference to work by Tomasello (2003), they give no grounds,
however, for this view of what syntax is. Nor do they explain how this view of
syntax could imbue their conclusion about the language of the MSA Blombos
inhabitants with the required pertinence.
The problem with this last conclusion is even more serious. Henshilwood
and d’Errico, as we saw above, have operated with a notion of ‘(syntactical)
language’ that is flawed in more than one way. Perhaps recognising this,
Henshilwood and Dubreuil (2009) adopt another notion of language:
The terminology used by archeologists to describe ‘modern’ language varies and
includes ‘syntactic’ language (Barham 2002), ‘symbolic’ language (Wurz 2000),

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 43

‘modern complex’ language (Mellars 2006), and ‘phonemic’ language (Klein and Elgar
2002). Broadly speaking, these terms are understood to mean the ideas or emotions that
were communicated by means of symbolic elements, for example vocally, by gesture,
or by marks, and that these elements can be recombined according to systematic,
conventionalized criteria to create meaning. Members of society interact with one
another in terms of their total culture through language, a non-instinctive, human
institution. It is according to the above definitions [sic] that the terms ‘modern’
language and ‘syntactic’ language are used in this chapter. (Henshilwood and
Dubreuil 2009: 41–2; emphasis added)

There are two fundamental problems with the notion of ‘modern’ or ‘syntactic
language’ assembled in this passage from the various ‘above definitions’. First,
it fails to draw a distinction between language as a means to communicate
ideas, emotions, etc. and the ideas, emotions, etc. communicated by means of
language. Evidently, what can be communicated by language does not form
part of language itself. Not drawing this distinction has odd consequences. One
instance should suffice; namely, that learning a language entails acquiring also
the ideas and emotions that can be communicated with the aid of it.
Second, if the technical terms ‘symbolic’, ‘syntactic’ and ‘phonemic’ mean
anything at all, then it is hard to see how the various ‘above definitions’ can
differ terminologically only. Much the same goes for ‘recursion’. If this tech-
nical term means anything, it is hard to see what it would mean to say that ‘the
Rubicon of modern language lies in recursion’. In their compound notion of
‘language’, Henshilwood and Dubreuil have conflated ‘symbolic language’,
‘syntactic language’ and ‘phonemic language’; and their aim in so doing may
well have been to accommodate the diverging definitions of ‘modern language’
adhered to by the archaeologists named in the quotation. The effect, though, is
to make it less clear still what entity the conclusion of the inferential step from
symbolic behaviour to fully syntactic language could be about. For instance, a
form of language or a stage in the evolution of language can be symbolic
without being syntactic; it is possible, after all, to make communicative use of
single linguistic symbols or syntactically unstructured strings of symbols. Then,
too, a form or a stage of language can be syntactic without being phonemic – if
it uses signing or writing for externalising utterances. Finally, such a form or
stage can be symbolic without being phonemic.
Henshilwood and his colleagues, then, are yet to come up with a solution to
the obscurity of their conclusion about the form of language used by the MSA
inhabitants of Blombos Cave. In the absence of such a solution, the conclusion
at issue cannot meet the Pertinence Condition. To solve the problem in a
satisfactory way, they need to underpin the conclusion with two well-justified
theories: a principled theory of what language is and an adequate modern
theory of syntax. As explained in Section 2.6, a theory of the former kind
is part of a more inclusive linguistic ontology. As for the theory of syntax,

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
44 Language Evolution

if a conclusion about language evolution is couched in terms that refer to


components of language structure such as syntax, morphology, phonology and
the lexicon, then it needs to be underpinned by theories that adequately
characterise these components in terms of principles, central properties and
subcomponents. In the same way, conclusions which express claims about
other aspects of language – for example, meaning, use, etc. – need to be
underpinned by theories – for example, a theory of semantics, pragmatics,
etc. – which give an adequate account of these aspects. Linguistic theories of
such components and aspects are members of a fifth kind of underpinning
theories, the other four kinds being bridge theories, grounding theories, a
linguistic ontology and a theory of processes of change that occur in the
linguistic domain.

3.4.3 The inferential step


Next, let us take up the inferential step shown in Figure 3.1 by the arrow
F. This step sets out from assumptions about one kind of thing – the
presumed symbolic behaviour of people who lived at Blombos Cave in
the Middle Stone Age. And it ends with a conclusion about a different kind
of thing – a particular form of language/syntax that these people used.
This gives rise to a question: ‘Why would it be permissible to infer from
the presumed symbolic behaviour of the MSA Blombos inhabitants that
they had fully syntactic or modern language?’ The question asks in effect
for an appropriate warrant for the inference. Doubts about the existence
of a simple link between symbolic behaviour and modern language/syntax
have been voiced by various archaeologists too, including Graves (1994),
Chase (1999), Mithen (1999), Davidson (2003, 2011), Coolidge and Wynn
(2011) and Wadley (2001).
As was made clear in Section 3.1, the warrant needed for an inferential step
takes the form of a bridge theory. Interestingly, the need for constructs such
as bridge theories has been expressly noted in some work that draws inferences
about the emergence of language or cognition from what is termed ‘primary
archaeological evidence’. For instance, d’Errico, Henshilwood and their
co-authors (2003) seem to be referring to just such constructs when they
speak of ‘general interpretive models’ (p. 51), of ‘analogies’ (p. 50) and of
‘frameworks of inferences that can establish a link between primary archaeo-
logical evidence and its wider implications’ (p. 54). But have they also used
bridge theories in attempts to warrant the inferential step from symbolic
meaning, behaviour or culture to the presence of modern or fully syntactic
language? And, if so, how adequate are those theories? We next take up the
two theories that have been prominent in the literature on the significance of
the Blombos finds.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 45

3.4.3.1 Transmitting and sharing symbolic meaning The bridge theory needed
for warranting the inferential step at issue is not to be had in any explicit terms
from earlier literature (Botha 2008a: 206–7, 2009b: 108–10). Only the core
assumption comes up, and it does so only in an allusive form, in claims such
as this:
The step from MSA symbolic meaning/behaviour to ‘fully
syntactical language’: a bridge assumption
Fully syntactical language is arguably an essential requisite to share and
transmit the symbolic meaning of bead works and abstract engravings
such as those from Blombos Cave. (Henshilwood et al. 2004: 404)
Similar claims in other places have it, in their various terms, that fully syntactic
language is ‘the only means of’, ‘essential for’ or a ‘direct link to’ the symbolic
meaning or behaviour at issue (e.g., d’Errico et al. 2003: 6; d’Errico et al.
2005: 19–20; Henshilwood and Marean 2003: 636). Other, but equivalent,
expressions such as ‘rely on’ have also been used in stating this assumption:
‘the transmission and sharing of the meaning of the engravings relied on fully
syntactical language’ (Henshilwood et al. 2002: 1279). The assumption at
issue is framed in related terms in more recent literature too:
Only a communication system like human language or equivalent to it can unambigu-
ously transmit the symbolic meaning of signs as well as the structured links
between them. (d’Errico and Vanhaeren 2009: 36)

The bare assumption expressed in the claim quoted above cannot, though, by
itself be the full bridge theory needed for warranting the inferential step at
issue. To be sufficiently well articulated, this theory – the ‘transmission
theory’, for short – needs to provide answers to questions such as these:
(i) What are the specifics of the meanings that were shared and transmitted?
(ii) For sharing and transmitting these specifics or specifics of this kind, why
was fully syntactic language needed rather than (syntactic) language at
some less fully evolved stage?
(iii) Why could these specifics not have been shared and transmitted by some
non-verbal means of communication?
(iv) How, in essence, do meanings whose transmission needs fully syntactic
language differ from meanings which are transmissible by less fully
evolved language or by non-verbal means?
(v) What are the cognitive capacities or processes that played a part in the
sharing and transmission of the meanings?
Questions (ii) and (iii) are particularly pertinent since, as we noted in Section
3.4.2, there is a widely held view that syntax evolved gradually in terms of
steps or stages.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
46 Language Evolution

To serve as a bridge theory, the transmission theory should meet the basic
conditions of adequacy that apply in empirical science. That is, it should
comprise hypotheses which are (i) testable, (ii) supported by empirical evi-
dence and (iii) non-ad hoc. In the absence of a bridge theory that meets these
conditions, the inferential step from symbolic meaning or behaviour to fully
syntactic language cannot be sound. Although Henshilwood and d’Errico do
not explicitly present a transmission theory which does fulfil these conditions,
it is possible that they draw on one that is implicit in the general archaeological
research paradigm to which they subscribe. Thus their view of the links among
MSA symbolic meanings seems to be similar in essence to views expressed by
Sally McBrearty and Alison Brooks (2000: 486), Paul Mellars (1998a: 95–6,
1998b: 89), Lyn Wadley (2001: 215) and others. That said, though, the early
literature seems not to hold any theory that explicitly addresses questions (i–v)
and that has been shown to meet the conditions of adequacy (i–iii). Nor does
one find such a transmission theory in recent literature in which it is main-
tained that
‘modern’ people must have had the facility for and used ‘syntactic’ language . . . as
without syntax in language it would arguably not have been possible to convey, within
and across individuals or groups, the meaning of these material symbols, for example
the rock art or personal ornaments. (Henshilwood and Dubreuil 2009: 42)

From the context, it is clear that, as used in these remarks, ‘syntactic language’
is synonymous with ‘fully syntactical language’.

3.4.3.2 Verbalising or articulating meta-representations The literature, then,


is lacking in transmission theories to warrant the step from symbolic mean-
ing/behaviour to modern or fully syntactic language. And Henshilwood and
Dubreuil seem to have given up on framing such a bridge theory. That is,
they (2009: 45–6, 51, 56) make the interesting admission that syntactic
language has been ‘linked in a questionable way to symbolic culture’. They
do so in response to Wynn and Coolidge’s (2007: 88) criticism that beads,
ochres and engraved bones cannot stand as evidence for modern cognition,
including language, unless the cognitive abilities needed for these artefacts
are spelled out. And, in proposing what the needed cognitive abilities may
be, Henshilwood and Dubreuil (2009: 55f.) put forward what may be looked
upon as a new bridge theory – the perspective-taking theory – to warrant the
inferential step from symbolic culture to syntactic language. In setting out
their argument, they make two terminological switches. One is from ‘sym-
bolic meaning’ and ‘symbolic behaviour’ to ‘symbolic culture’. The other is
from ‘fully syntactical language’ to ‘syntactic language’. Neither switch
materially affects the question of the soundness of the inferential step shown
as F in Figure 3.1.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 47

In outline, Henshilwood and Dubreuil (2009) reason as follows to warrant


the inferential step from MSA symbolic culture to modern language:

The step from MSA symbolic culture to modern language:


a bridge theory
Claim 1: The invention and maintenance of MSA symbolic culture –
such as that attributed to the Blombos inhabitants – requires the
higher-level cognitive ability of level-2 perspective-taking.
Claim 2: Level-2 perspective-taking involves the construction of
hierarchical meta-representations.
Claim 3: Recursive syntax is required for verbalising or articulating
these meta-representations.
Claim 4: Recursive syntax is the defining feature of modern language.
Conclusion: Modern language, therefore, ‘came along’ with level-2
perspective-taking as part of modern cognition.
Let us consider these claims each in turn. As regards Claim 1, level-2
perspective-taking is said to be the ability to represent in the mind different –
and even conflicting – perspectives of an object. This involves the ability to
reconstruct how an object looks from another person’s perspective. It is claimed
to appear in children at a relatively young age as the result of a marked
change in their cognition. Henshilwood and Dubreuil now contend that
the emergence of artifacts such as Blombos beads and engraved ochres could indicate
that a similar cognitive change occurred during the MSA. Beads could come to
symbolize social statuses (e.g. one’s position within a kinship structure), because people
would have been able to recognize the stability of its meaning across contexts and
perspectives [emphases added]. (Henshilwood and Dubreuil 2009: 55)

It is clear from the words emphasised that, for claims made in empirical work,
these claims are quite speculative. But let us accept for the sake of argument
that the humans at issue did have the ability of level-2 perspective-taking. Let
us also accept Claim 2, that is, that level-2 perspective-taking involves the
construction of hierarchical meta-representations, meta-representations being
higher-order representations of representations.
We turn next to Claim 3, a claim crucial to Henshilwood and Dubreuil’s
reasoning about the way in which syntax is linked to the meta-representations
at issue. Their unpacking of this claim is instructive:
Imagine a language with a linear syntax, in which the meaning of a word changes with
the position of the word in the sentence. The meaning of ‘Bob hit Fred,’ for instance,
would be different from the meaning of ‘Fred hit Bob.’ Such a language would be
insufficient to verbalize the kind of meta-representations associated with level-2
perspective-taking and ToM [i.e., theory of mind]. Meta-representations have to
be articulated in a hierarchical way by embedding clauses, as in sentences like

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
48 Language Evolution

‘Fred sees that I wear the beads’ or ‘Fred knows that I am the chief.’ Without
recursive syntax, it is impossible to articulate conflicting perspectives.
(Henshilwood and Dubreuil 2009: 59)
Though the expressions ‘verbalize’ and ‘articulate’ are crucial in Claim 3, their
meaning is unclear, admitting of at least two construals. First, to say that clausal
embedding is needed ‘to verbalize’ or ‘articulate’ a meta-representation may be
to claim that conveying the content of the meta-representation needs a particular
kind of complex sentence. Such a claim would be false, however, since any unit
of (cognitive) content or (semantic) meaning can be conveyed by using non-
complex sentences, that is, clauses, in juxtaposition. Neither the whole nor the
components of such sentences are formed by recursion. This has been illus-
trated by Newmeyer (2004: 4). He observes that the propositionally complex
meaning conveyed by a complex sentence such as ‘Mary thought that John
would leave.’ can be conveyed just as fully by juxtaposition of two clauses such
as those in ‘Here is what Mary thought. John was going to leave.’ Similarly, to
verbalize or articulate the content of a particular meta-representation, one can
use either a complex sentence, say ‘Fred sees that I wear the beads’, or a
juxtaposition of two clauses, such as ‘Fred sees this. I wear the beads.’
The clauses can even be used in the reverse order – that is, ‘I wear the beads.
Fred sees this’ – to convey the complex meaning in question.
Independent evidence is to hand that complex sentences are not needed to
convey complex meanings of the sort at issue. Thus, on various analyses, there
are restricted linguistic systems that make no use of clausal embedding. These
include (i) the systems acquired by adults who learn a second language
naturally (Klein and Perdue 1992, 1997), (ii) homesign systems created by
deaf children of hearing parents (Heine and Kuteva 2007: 202), and (iii) twins’
languages (Bakker 1987, 2006; Heine and Kuteva 2007: 202). In addition,
early pidgins and creoles are claimed to have less subordination than their later
developmental forms (Heine and Kuteva 2007: 287ff.). And, on some ana-
lyses, there are full modern languages – for example, Pirahã (Everett 2005,
2009; Parker 2005) – which use recursion in a limited way, if at all.11 That
languages only need quite limited syntactic means for the task of communi-
cation is also shown by David Gil’s (2005, 2008a, 2010) analysis of Riau
Indonesian, a point to be fleshed out in Section 3.4.3.3 below.
Getting back to the perspective-taking bridge theory, since Claim 3 is
untenable, it does not follow that the Conclusion is true. That is, it does not
follow that modern language came along with level-2 perspective-taking
as part of modern cognition. This conclusion would not even follow if
Claim 4 – namely, that recursive syntax is the defining feature of modern
language – were accepted. The latter claim represents an assumption which has
probably been derived from an account by Hauser et al. (2002), and is itself
controversial (e.g., Everett 2005, 2009; Parker 2005), a point which will be

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 49

taken up again in Section 11.4 below. In its present form, then, Henshilwood
and Dubreuil’s account of the link between level-2 perspective-taking and
recursive syntax falls short of providing the bridge theory needed to warrant
the inferential step from symbolic culture or behaviour to fully syntactic or
modern language. And, tellingly, Henshilwood and Dubreuil concede this in
their recent review article, stating that
[i]n Henshilwood and Dubreuil (2009: 59), we claim that syntactic recursion is essential
to express metarepresentations. Botha (2010) argues that this is not necessarily the case,
a point also granted in Dubreuil (2010a: 124). (Henshilwood and Dubreuil 2011: 392)

The upshot is that at this stage Henshilwood, d’Errico and Dubreuil have no
bridge theory to warrant the inferential step at issue. First, then, this step fails
the Warrantedness Condition and is therefore unsound. Second, the conclusion
that the MSA inhabitants had fully syntactic language is incorrect. This second
point, to be sure, does not imply that those humans did in fact lack fully
syntactic or modern language. But it does have a restrictive implication.
It implies that, without drawing on an adequate bridge theory, amongst other
necessities, it is not possible to infer from assumptions about the presumed
symbolic culture or behaviour of these humans that they had fully syntactic or
modern language. Henshilwood and Dubreuil (2011: 392) do not have a new
bridge theory to offer in place of the untenable transmission theory and
collapsed perspective-taking theory. And in response to criticism by Davidson
(2011: 382–3), they even go so far as to concede that they are ‘shift[ing] away
from syntax’. The transmission theory and the perspective theory fail as bridge
theories for fundamental linguistic reasons, to which we turn in the next section.

3.4.3.3 Linguistic cause of the weakness of the bridge theories Both the trans-
mission theory and the perspective-taking theory proceed from the following
assumption:
Assumption about expressive power of simple linguistic means
Complex meanings or units of cognitive content cannot be transmitted/
shared/articulated with the aid of relatively simple linguistic means.
Underlying this assumption, there is too low an estimate of the expressive
power of relatively simple linguistic means such as parataxis. This wrong view
may in turn be due to a more general misunderstanding of how the complexity
of language relates to that of a phenomenon such as symbolic meaning,
behaviour or culture. That relation is not a simple one in the sense that an
increase in the complexity of some non-language factor forces a rise in the
complexity of language. In particular, as has been shown in Section 3.4.3.2,
what are presumed to be complex meanings or units of cognitive content can
be expressed by relatively simple syntactic means.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
50 Language Evolution

Some version of the assumption at issue appears to be implicit in the view


that modern language is needed for the generation or creation of particular
kinds of symbolic codes. Consider in this regard the following claims:
Modern language is the only communication system with a ‘built-in’ metalanguage that
allows the creation of symbolic codes. (Henshilwood 2007: 123)
Human language is the only known natural system of communication that has a built-in
metalanguage that enables the generation of other hierarchically structured
symbolic codes. (d’Errico and Vanhaeren 2009: 36)
It is not explained, though, why language would have had to be syntactically
complex – incorporating recursion, for instance – to fulfil the metafunction in
question. Or, equivalently, why the symbolic codes in question – no details of
which are given – could not have been created with the aid of simpler linguistic
means forming part of stages of language that preceded modern language.
The untenability of the assumption about the limited expressive power of
relatively simple linguistic means shows up starkly in Gil’s (2008a, 2010)
analysis of Riau Indonesian. (This is one of the colloquial varieties of Malay/
Indonesian, just one from a wide range.) On the analysis by Gil, the language
has a very simple grammar, yet adequately fulfils the functions of a complex
society. That is, he maintains that Riau Indonesian’s quite limited morpho-
logical, syntactic and semantic means are ‘enough to run a community of some
hundred million people, and, by extension, most contemporary human activity
throughout the world’ (2008a: 127, 2010: 30).
As for the linguistic simplicity of Riau Indonesian, it is what is known as
a Relative I(solating)-M(onocategorial)-A(ssociational) Language; that is, a
language whose properties closely resemble those of a Pure IMA Language.
Gil (2008a: 124, 2010: 20) assigns three properties to a Pure IMA Language:
Properties of a pure IMA language
(a) It is morphologically isolating in that it has no word-internal morpho-
logical structure. For instance, it has no inflected, derived or compound
words.
(b) It is syntactically mono-categorial in that it has no distinct parts of
speech such as nouns and verbs.
(c) It is semantically associational in that it has no construction-specific
rules of semantic interpretation.
Riau Indonesian is viewed by Gil as a Relative IMA Language since it is slightly
more complex than a Pure IMA Language in a number of ways (none of which
is material to the discussion). At the same time, though, it is far less complex
grammatically than languages such as English and Hebrew. Thus, on Gil’s
(2008a: 126, 2010: 23) analysis of Riau Indonesian, it does not have distinct
open syntactic categories (or parts of speech) such as verbs and adjectives.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 51

For instance, the two ‘content words’ that make up the non-telegraphic sen-
tence Ayam makan, ‘Chicken eat’, are both members of the single open
category ‘sentence’. And so Ayam makan is a simple juxtaposition or coordin-
ation of two sentences. The sentence has the vague meaning of ‘something to
do with chicken and eating’; this can, however, be made more specific by the
context in which it is uttered. As Gil (2008b: 23) puts it, speakers ‘exercise
their pragmatic competence and plumb the available linguistic and extra-
linguistic context in order to fill in as much detail as is considered necessary’.
As a result, on Gil’s (2010: 23) account, utterances of the sentence Ayam
makan can be understood in many different ways, among them ‘The chicken is
eating’, ‘The chickens that were eaten’ and ‘The reasons chickens eat’.
According to Gil, it is possible in Riau Indonesian to construct longer and
more complex sentences that have the same kind of simple IMA structure as
Ayam makan. This means, amongst other things, that such sentences are
constructed without the aid of the kind of recursion that Henshilwood and
Dubreuil take to be the defining feature of modern language.12
The grammatical simplicity of Riau Indonesian, interestingly, does not
cause the language to lack expressive power in any semantic domain
(Gil 2008a: 127, 2010: 30). As Gil sees it, this state of affairs goes to show
that linguistic complexity is not needed to support complexity in domains such
as culture, technology and civilisation (Gil 2008a: 123). What is more, the fact
that Riau Indonesian is able to meet the communicative and other needs of a
contemporary community by relatively simple linguistic means prompts him
to come up with a radical suggestion that
no amount of non-linguistic, archeological evidence will ever be able to prove the
presence, at some earlier stage of human evolution, of grammatical entities of greater-
than-IMA complexity. (Gil 2008a: 128, 2010: 31)

3.5 Conclusion
This chapter is concerned with the soundness of the Blombos inference and
with the heuristic power of the associated shell-bead window on language
evolution. What, then, can be learned about these two matters from the
analysis set out in the previous sections?
To recapitulate briefly, then. The first step in the Blombos inference – that
from shells to beads – is not problematic, as far as I can judge. It starts out from
a body of data that has been collected, analysed and checked in a meticulous
way. The step is, therefore, well grounded. The conclusion with which the step
ends up – that the shells were worn as beads or components of bead works – is
not contentious from the viewpoint of pertinence; it is clear, that is, what beads
are in the particular context. And the inferential step from the data about
the shells to the conclusion assigning beadhood to them has a warrant that

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
52 Language Evolution

does not seem to be problematic. That is to say, the step is underpinned


by bridge assumptions which do not appear to be controversial.
The second inferential step – that from beads to symbols/symbolic
behaviour – is much less fully articulated than the first one. This goes for the
data or assumptions about beads in which the step needs to be grounded. It
goes, too, for the warrant needed to license the move to the conclusion that the
beads had the status of symbols. The conclusion, what is more, has been
criticised as falling short of pertinence on the grounds that the notion of a
‘symbol’ has been left insufficiently well defined. Consequently, it has been
argued, two things are unclear: it is unclear what the distinctive properties of
symbols are, and it is unclear how symbols are to be identified. Lately,
the notion of a ‘symbol’ adopted by Henshilwood and Dubreuil (2011) has
been questioned again, and from various perspectives, by Coolidge and
Wynn (2011: 380–1), Malafouris (2011: 385), Gärdenfors (2011: 383),
and Wadley (2011: 388–9). The inferential step from beads to symbols is
contentious, in short, from the viewpoint of all three soundness conditions.
The third inferential step – from symbols/symbolic behaviour to modern/
fully syntactic language – is the most problematic of the three. To start with, it
is insufficiently well grounded owing to the paucity of specific data about the
symbolic behaviour of the MSA inhabitants of Blombos Cave. Next, its
conclusion is phrased in terms of an ill-defined notion of ‘modern language’
or ‘fully syntactical language’. Finally, its move from data to conclusion lacks
a warrant that is underpinned by an adequate bridge theory.
Being a compound inference, the Blombos inference cannot be sound unless
each of its three steps is sound. It fails to meet that requirement, though;
clearly, the step from symbolic behaviour/culture to modern/fully syntactical
language is not sound. This finding is echoed in the following remarks by
Coolidge and Wynn:
We do deny that shell beads are prima facie evidence for modern syntactical language.
Botha (2010) has made clear the dangers of this inferential leap. Beads were made
intentionally. Yes. Beads were worn. Yes. Beads may have some personal or limited
interpersonal (collectively shared) meanings. Yes. Beads are evidence of modern
syntactical language. No! Not necessarily. Taking shell beads as evidence of syntactical
language is an unjustified inferential leap and the authors’ [i.e., Henshilwood and
Dubreuil’s] implication that full behavioral modernity is attached to shell bead making
and incised ochre is unjustified as well. (Coolidge and Wynn 2011: 381)
As we noted in Section 3.1, the Blombos inference is representative of a whole
category of archaeological inferences about language evolution. The problems
with the soundness of this inference may reflect, accordingly, on this category
as a whole. And they do. The problem(s) with the symbolhood of shell beads;
the lack of specificity about what MSA symbolic behaviour involved; the
obscurity of the entity denoted by ‘fully syntactical language’ or ‘modern

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
Sea shells, ancient beads and Middle Stone Age symbols 53

language’; and the absence of an adequate bridge theory warranting the


inferential step from symbols/symbolic behaviour to fully syntactical/modern
language – these flaws are not unique to the Blombos inference. On the
contrary, they represent limitations of the category of all inferences that start
out from data about properties of artefacts claimed to have symbolic meaning,
and move stepwise from there to a conclusion which ascribes modern language
to some group of prehistoric humans. The category of such inferences is
potentially a large one. This is due to the variety of the artefacts at issue; as
well as shell beads, they include engraved ochres, engraved or decorated bone
tools, funerary goods and so on.13 MSA shell beads have been recovered, what
is more, at sites other than Blombos Cave; these include Skuhl and Qafzeh in
Israel, Oued Djebbana in Algeria, Grotte des Pigeons in Morocco and Sibudu
Cave in South Africa.14
This brings us to the question of the heuristic power of the window on
language evolution that provides the conceptual means for drawing the Blom-
bos and similar other inferences. Specifically: what do the problems with the
soundness of these inferences reveal about the capacity of this window to yield
interesting insights into the evolution of language? For instance, the potential
insight expressed by the conclusion that modern language was used as early as
75,000 years ago on the south-western coast of South Africa. The problems at
issue indicate that the conceptual means provided by the shell-bead window is
insufficiently well developed. It is in more than one way still a ‘window under
construction’. Substantial additional work is needed, for example, to construct
a theory of what MSA symbolic behaviour involved as well as a bridge theory
to warrant the inferential step from symbolic behaviour to fully syntactic
or modern language. In addition, the notion of ‘fully syntactic language’ or
‘modern language’ needs to be underpinned by principled theories of what
language and syntax are. Unless the further work on the window – and in
particular the construction of the theories needed to underpin components of
the window – is done in a principled way, the window will remain underdevel-
oped. This requirement means, for instance, that one cannot simply stipulate
that some property of language – for example, recursion – is the defining
feature of modern language. And one cannot simply abandon this position later
without further ado when it is shown to have unfortunate consequences, even
going so far as to ‘shift away from syntax’ as not being ‘a cognitive develop-
ment of its own’ (Henshilwood and Dubreuil 2011: 392). To do this is to create
an impression of arbitrary conduct; the impression that one has decided on a
priori grounds that the MSA inhabitants of Blombos Cave had modern lan-
guage; and that one is willing to embrace any linguistic theory, regardless of its
adequacy, if it could be said to support this decision.
In closing, let us return to d’Errico’s view that archaeologists and palaeoan-
thropologists ‘probably retain in their hands the best, not to say the only,

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004
54 Language Evolution

information’ to answer the question of ‘what syntactical language is and where


it arose’ (in Számadó et al. 2009: 231). What should one make of this view
against the background of the discussion above? At first blush, there are two
possibilities. First possibility: d’Errico has got it right. But what would this
mean? Suppose that there is not any better or other ‘information’ about what
syntactic language is, and where it arose, than that which archaeologists
and palaeoanthropologists have. Suppose, too, that the ‘information’ about
MSA shell beads and similar other artefacts is the best that archaeologists and
palaeoanthropologists have. Then, in view of the unsoundness the Blombos
inference and of similar other inferences, it would not be possible to learn
anything about what syntactic language is and where it arose. An unattractive
possibility indeed. Second possibility: d’Errico has got it wrong. That is to say,
his view represents an overestimation of the value of the ‘information’ which
archaeologists and palaeoanthropologists have about what syntactic language
is and where it arose. There is in fact a third possibility, though. It is more
attractive than either of the first two, since it is the possibility that there may
be no need to choose between them. For, on more careful consideration,
d’Errico’s view of the value of the ‘information’ at issue is seen to miss a
crucial point which emerged from the discussion in earlier sections; that is,
raw ‘information’ does not speak for itself. The value of such ‘information’
depends, instead, on the adequacy of the conceptual means used in
drawing inferences from it. And the inadequacy of the means in use at present
makes it impossible to judge the value of ‘information’ such as that about
MSA shell beads.

Downloaded from https://www.cambridge.org/core. University of New England, on 07 Nov 2017 at 01:55:41, subject to the Cambridge Core
terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781316471449.004

You might also like