992 LANGUAGE, VOLUME 92, NUMBER 4 (2016)

speech communities can undergo very rapid language change, especially if there is no braking ef-
fect from widespread literacy in the language or the use of the language in school and official
contexts. Ideally, authors would describe the target language spoken around the children and not,
for example, data gathered thirty years earlier. Some authors source target-language examples
from standard reference grammars, but not all authors do, leading to examples with spellings or
glosses that differ considerably from the recorded grammatical descriptions (e.g. Warlpiri exx. 2,
3a–c, 4b, 13); are they errors or hitherto unattested forms/glosses? It is crucial that each type of
data be properly sourced, including the time of gathering.
Studying the acquisition of language in small speech communities is not easy—it requires
mastery of the target language, good understanding of the context, and excellent relations with
caregivers. These authors provide rich material for research into acquisition of language in small
communities and for understanding ergativity.

REFERENCES
O’Shannessy, Carmel. 2006. Language contact and children’s bilingual acquisition: Learning a mixed lan-
guage and Warlpiri in northern Australia. Sydney: University of Sydney dissertation. Online: http://hdl
.handle.net/2123/1303.
Stoll, Sabine, and Balthasar Bickel. 2012. How to measure frequency? Different ways of counting erga-
tives in Chintang (Tibeto-Burman, Nepal) and their implications. Potentials of language documenta-
tion: Methods, analyses, and utilization (Language documentation & conservation special publication
3), ed. by Frank Seifart, Geoffrey Haig, Nikolaus P. Himmelmann, Dagmar Jung, Anna Margetts, and
Peter Trilsbeek, 83–89. Honolulu: University of Hawai‘i Press. Online: http://hdl.handle.net/10125
/4520.

ARC Centre of Excellence for the Dynamics of Language

Australian National University, Canberra, Australia
[[email protected]]

Why only us? Language and evolution. By Robert C. Berwick and Noam Chomsky.
Cambridge, MA: MIT Press, 2016. Pp. 224. ISBN 9780262034241. $22.95 (Hb).
Reviewed by Ljiljana Progovac, Wayne State University
1. Chomsky’s knot. It is a good thing that we linguists are discussing language evolution, for
without linguists’ input, this important question cannot be properly addressed. Likewise, ‘like
other biological phenomena, language cannot be fully understood without reference to its evolu-
tion, whether proven or hypothesized’ (Givón 2002:39, emphasis added). In that sense,
Berwick and Chomsky’s book is important; being written by scholars of such great stature, it
sends a clear message that this topic is both timely and relevant for linguists. While there has
been a slight (tacit) shift from some of their previous claims, my conclusion is that B&C’s pro-
posal still keeps them and many other linguists tied in a knot, a knot that prevents them from de-
veloping new hypotheses and angles to be explored. This leaves ample room for the alternative
approaches to language evolution, the ones that B&C dismiss.
It is encouraging to see that B&C have softened, as least to some extent, their original stance
on the emergence of language, as well as the vehemence of their criticism of opposing views. For
example, while they do not acknowledge this, B&C have significantly shifted their estimated date
of the emergence of language to up to 200,000 years ago (157), from the previous ‘just a bit over
50,000 years ago’ (Chomsky 2005). In this respect, they meet almost half way Dediu and Levin-
son’s (2013) estimate that language dates back to the common ancestor of humans and Nean-
derthals, to some 400,000–500,000 years ago (for sharp criticism of Dediu and Levinson’s
claims, see e.g. Berwick et al. 2013). Not only that, but B&C no longer claim that Neanderthals
did not have language. Instead, B&C now say that it is the ‘$64,000 question whether Neander-

Printed with the permission of Ljiljana Progovac. © 2016.

 REVIEWS 993

tals had language’ (50). This shift comes in the midst of a host of very recent findings in both
archeology and genetics that point to a deeper timeline for the emergence of language. Such find-
ings will continue to accrue and surprise us.
In this short review, I can only address a few claims among many made by B&C. I see two main
threads streaming into what I call chomsky’s knot (on analogy with B&C’s references to ‘Dar-
win’s problem’ or ‘Darwin’s troubles’). The first thread is their claim that the only serious way to
approach the question of language and its evolution is to adopt the most recent theoretical postu-
lates of Chomsky’s framework minimalism (in particular, the strong minimalist thesis), which re-
duces syntax to a single optimal operation Merge: ‘UG [universal grammar] must meet the condi-
tion of evolvability, and the more complex its character, the greater the burden on some future
account’ of its evolution (93). The second thread is the claim that ‘we simply do not have as much
to explain’ (11): given how simple syntax must be,1 the evolution of syntax/language amounted to
just one single, unremarkable event. In other words, according to B&C, (i) in order for syntax to be
evolvable, syntax itself has to be extremely simple, and (ii) given that syntax must be super simple
(as per (i)), syntax must have arisen through one single, minor mutation. This proposal is circular
and entangled. It has kept many researchers tied in a knot, since this reasoning makes unclear what
remains to be learned either about syntax or about the evolution of syntax.
2. Does merge combine CONCEPTS? merge, move, perfection, and more. B&C simply as-
sume that there is in fact ‘the optimal situation’, contra appearances (of great complexity), and
that ‘UG reduces to the simplest computational principles … This conjecture is sometimes called
the Strong Minimalist Thesis (SMT)’ (94). The claim is that the ‘generative process is optimal’,
based on ‘efficient computation’ (71), and that ‘this newly emerged computational system for
thought … is perfect, in so far as SMT is correct’ (80). However, B&C give no definition of ‘op-
timal’ or ‘efficient’ or ‘perfect’. And they do not know if the SMT is correct (see below).
Evoking the SMT, B&C define Merge as ‘the simplest possible mode of recursive generation:
an operation that takes two objects … and forms from them a new object … the set’ (70). But
what are these ‘objects’ that get combined by Merge? B&C claim that Merge ‘takes human con-
cepts as computational atoms and yields structured expressions that … provide a rich language
of thought’, suggesting that ‘these processes might be computationally perfect’ (87, emphasis
mine). But what exactly are ‘human concepts’, and what exactly is ‘perfect’ in this context? Cru-
cially, syntactic Merge, or any other combinatorial operation that goes by different names in dif-
ferent linguistic frameworks, is taken by linguists to combine words, and not units of thought.2
Linguists have at least converged on reasonably reliable definitions and characterizations of what
a word is, or what a morpheme is, but there are no comparable characterizations of what a human
concept would be, at least none that are introduced in B&C’s book.
B&C in fact state that ‘the atomic elements [of Merge] pose deep mysteries’. They are ‘word-
like, but not words … Their origin is entirely obscure’ (90). However, claiming that the purpose
of Merge is to combine something that we have no characterization of is not very helpful. For if
we do not know what the units of Merge are, how can we possibly know or even hypothesize
about what cognitive and reproductive advantages resulted from combining them? Did B&C con-
clude that the only way to continue to appeal to mathematical purity and perfection, which they
seek, is to replace messy and imperfect words with something intangible, something that can at
least in theory be ‘perfect’? According to B&C, to understand evolution ‘requires a more subtle
mathematical analysis, and so far as we can make out, none of the recent books on the evolution
of language seem to have grasped this in full’ (16). In fact, B&C suggest that Darwin did not ei-
ther, since he was not mathematically minded, and they quote from Darwin’s autobiography:

1 Contra appearances, as linguists from a variety of frameworks and language backgrounds have found oth-

erwise. B&C maintain that this appearance of complexity and diversity simply reflects ‘a lack of deeper un-
derstanding’ (93).
2 The general understanding in syntactic analysis is that words have various (idiosyncratic) grammatical

features that drive Merge in the first place and determine what is possible to Merge and what is not. For an ac-
cessible introduction to this framework, the reader is referred to, for example, Adger 2003.
994 LANGUAGE, VOLUME 92, NUMBER 4 (2016)

‘ “my power to follow a long and purely abstract train of thought is very limited; and therefore I
could never have succeeded with metaphysics or mathematics” ’ (16).
More intriguingly yet, given B&C’s vague proposal, how do we prove or disprove that some
(or all) other animals do not have this same kind of thoughtful (and unobservable) Merge, and
with it, thus, the essence of language and thinking? As B&C themselves observe, ‘we know that
non-human animals excel at many challenging cognitive tasks’, such as making tools and causal
reasoning (139–40).
On p. 74, there is some discussion in relation to so-called Internal Merge (formerly called
Move), meant to illustrate the workings of a computationally perfect system. As B&C show,
Move applies to John is eating what to yield the question What is John eating what?, from which
the initial copy of what is deleted (not pronounced). According to B&C, ‘to pronounce more
[copies of Move] would yield enormous computational complexity’ (101). First of all, it is not
clear why pronouncing copies that have already been created by grammar would increase com-
putational cost, let alone enormously. If anything, pronouncing or deleting a copy looks like one
of those surfacey ‘externalization’ phenomena that B&C have deemed irrelevant for syntax
proper. And why does the very application of Move, a significant distortion of the original sen-
tence structure, not incur more computational cost than a mere pronunciation of a copy already
created? This question becomes that much more relevant in the light of the fact that in some lan-
guages, such as Japanese and Chinese, there is no (visible) Move of wh-phrases, resulting in
questions of the kind John reads what?.
B&C further claim (74) that it is the final copy of Move that has to be pronounced (rather than
initial), for ‘otherwise, there will be no indication that the operation has applied’ (emphasis
mine).3 But why does there have to be an overt (externalized) indication that the operation has
applied, unless this is for the purposes of communication? Why would the language of internal
thought care about this? Astonishingly, however, B&C proclaim that this property of syntax, to
pronounce only one (final) copy, demonstrates that ‘language evolved as an instrument of
thought, with externalization [communication] a secondary process’ (74).
Finally, it is rather curious to find that B&C do not in fact believe that the SMT, the cornerstone
of their proposal, has been established, or that it necessarily will be. They state that ‘some years
ago, SMT would have seemed a very exotic idea. But in recent years evidence has been accumu-
lating suggesting that something like this may hold considerable promise. That would be
a surprising and significant discovery if it can be established’ (94, emphasis mine), but ‘the
SMT is very far from established’ (71). In fact, this proposal reads more like a legal document
than a scientific proposal: it hedges and avoids explicit claims, and is lacking in specific hy-
potheses that can be subjected to verification and falsification.
3. What’s wrong with natural selection, gradualism, and darwin? Ch. 1 covers a
host of new claims and findings in the field of evolutionary biology, with one goal being to dis-
credit all of the recent books on language evolution. These, according to B&C, put too much em-
phasis on natural selection, which they argue faces serious problems.4 However, given their
blanket criticism of all of the recent books on language evolution, but lacking any specific details,
it is not really possible or advisable to respond.5

3 If so, then Japanese and Chinese wh-questions would need to be analyzed as lacking Move, rather than as

pronouncing the initial copy, as some syntacticians have proposed.

4 There is imbalance in this book between covering evolutionary biology vs. syntactic theory. B&C proceed

as if minimalism is a given, well known, and hardly controversial, while Darwin’s natural selection is in seri-
ous trouble.
5 In fact, as far as B&C are concerned, these recent books on language evolution should not exist at all.

B&C’s book is full of warnings and reprimands aimed at these books, seen as clear undesirables. Their au-
thors were supposed to heed Lewontin’s (1998) warning that the question of the evolution of language and
cognition is almost impossible to address, and when they did address it, it was to their own detriment (B&C,
p. 97). But B&C have obviously written books and articles about language evolution. Are B&C trying to say
that this incredibly hard topic is only for a select few: B&C and a few close followers?
 REVIEWS 995

First, B&C point out that natural selection for beneficial traits is extremely hard to achieve
(21–22), since there are typically also opposing forces working against it, such as genetic drift or
chance. On p. 59 B&C do acknowledge that Darwin himself never thought that natural selection
for a beneficial trait was the only or the exclusive mechanism of evolution. Nonetheless, for all
relevant intents and purposes, natural selection in the Darwinian fashion does happen, as the au-
thors themselves acknowledge (26) and actually adopt in later chapters (e.g. p. 59). So, perhaps
the message of Ch. 1 is not that one should never invoke natural selection when it comes to lan-
guage evolution, but that one should not appeal to it too much. Elsewhere in the book, however,
as well as in previous work, B&C suggest that appealing to natural selection via tinkering can be
symptomatic of the lack of understanding: ‘if you take a look at anything that you don’t under-
stand, it’s going to look like tinkering’, but when things are properly understood, one realizes that
there is much more order in nature (Chomsky 2002:139).
B&C point out that it is especially difficult to spread a completely novel mutation, before it
reaches some critical ‘tipping point’ in the population (80). While B&C seem to claim that one
single novel mutation emerged in humans and gave them Merge (and with it syntax and the in-
ternal language of thought) and that this single mutation got naturally selected (because it was
beneficial for thinking and planning), it need not be the case that the ability to use syntax or lan-
guage in general had to wait for a completely novel mutation to emerge. It is entirely possible that
the initial (simpler) stages of language recruited elements of existing genetic makeup, possibly
clusters of mutations, through the selection of those individuals who were just a little better at
combining words or storing words in the memory. In fact, the problem for natural selection that
B&C discuss is much more of a problem for their own abrupt and discontinuous approach, and
much less of a problem for a gradualist, step-by-step approach.6
In Progovac 2015, I have proposed such a gradualist approach to the evolution of syntax, argu-
ing, contrary to B&C, that one can decompose the attested complexities of syntax into evolution-
ary stages/primitives, with each new stage leaning on, and overlapping with, the previous one(s)
and bringing about some concrete, incremental communicative benefits.7 Utilizing certain stable
theoretical postulates of minimalism and its predecessors, this approach reconstructs the initial
stage of syntax as a flat, intransitive, absolutive-like, two-word stage, which does not manifest a
distinction between subjects and objects. This is arrived at by peeling off the layers of sentential
hierarchical structure (tense phrase/TP and transitive verb phrase/vP), but also by taking language
variation into account. This foundational stage is seen as the common denominator for the at-
tested variability in, for example, the expression of transitivity across languages and construc-
tions, including nominative-accusative, ergative-absolutive, and serial verb patterns. Identifying
such a basic stage also goes a long way toward revealing some points of contact and continuity
with the capabilities of other species. For continuity is not to be sought in the most complex of
grammatical abilities, but rather in the simplest of syntax.
On this approach, one need not deny the ‘appearance’ that today’s syntax is rich and complex
or that syntactic variation across languages is significant, but can actually use both to formulate
concrete hypotheses about language evolution, specific enough to engage genetics, neuroscience,
and the hominin timeline. This gradualist approach views syntax less as a precise mathematical
formula and more like a quilt, stitched together from a variety of patterns and structures, accrued
at various junctures in language evolution. This is also how geneticists describe the human
genome: as ‘a patchwork quilt … with segments that were picked up at different stages of our an-
cestry’ (Harris 2015:xvii). It may seem ‘messy’ to some, but there is pattern and richness in this
quilt of syntax, as well as a plenitude of clues about its evolutionary trajectory.
4. The importance of the scientific method. B&C put forth some sensationalist claims, in-
cluding that sophisticated Martian scientists would consider all of the organisms on Earth as one
6 The novel mutation scenario is preferred by B&C because they insist on a total discontinuity with other

species when it comes to the capacity for language. If the initial selection targeted mutations that were already
available in some individuals of the other species, then the divide between ‘us and them’ cannot be as sharp
as B&C envision. But B&C do acknowledge on p. 52 that, in principle, selection can make use of variation al-
ready present in a population.
7 [Editor’s note] Progovac 2015 is also reviewed in this issue of Language.
996 LANGUAGE, VOLUME 92, NUMBER 4 (2016)

and the same (61) and all human languages as one and the same (78), and that an infant from a
Stone Age tribe in the Amazon would, if brought to [today’s] Boston, be indistinguishable in lin-
guistic and other cognitive functions from the rest of the children (54). This latter claim comes
from their view that all humans are identical in their genetic basis for language (with the excep-
tion of language disorders) and that humans have always had this same genetic basis, even in pre-
history. The burden has to be on B&C to at least identify a method by which one can prove or
disprove claims of this kind.
Suffice it to say here that recent research has found there to be individual linguistic differences
that correlate with genetic differences, even among healthy adults, contra B&C’s claim on p. 55.
This is, of course, especially evident with disorders. In the very last footnote of their book, B&C
do acknowledge that there may indeed exist ‘some language variation in “normal” human popu-
lations that is being uncovered by genome sequencing’ (177). They quote Kos and colleagues
(2012), who found that CNTNAP2 gene SNP variants in human populations affected language
processing in otherwise normal adults. At the population level, researchers have been looking
into possible correlations between parameters of language variation and genetic makeup, starting
with the pioneering work by Dediu and Ladd (2007).
In conclusion, contrary to B&C’s claims, there is an enormous amount to discover when it
comes to language evolution, and the only way to proceed is to consider a variety of hypotheses,
to subject them to testing and falsification, and to emerge from this process with ever better and
more refined hypotheses. Simply making claims about things and hoping to be right is not nearly
as useful as generating specific and testable hypotheses, even if they turn out to be wrong.

REFERENCES
Adger, David. 2003. Core syntax: A minimalist approach. Oxford: Oxford University Press.
Berwick, Robert C.; Marc D. Hauser; and Ian Tattersall. 2013. Neanderthal language? Just-so stories
take center stage. Frontiers in Psychology 4:671. DOI: 10.3389/fpsyg.2013.00671.
Chomsky, Noam. 2002. On nature and language, ed. by Adriana Belletti and Luigi Rizzi. Cambridge: Cam-
bridge University Press.
Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36.1–22. DOI: 10.1162/00243
89052993655.
Dediu, Dan, and D. Robert Ladd. 2007. Linguistic tone is related to the population frequency of the adap-
tive haplogroups of two brain size genes, ASPM and Microcephalin. Proceedings of the National Acad-
emy of Sciences 104.10944–49. DOI: 10.1073/pnas.0610848104.
Dediu, Dan, and Stephen C. Levinson. 2013. On the antiquity of language: The reinterpretation of Nean-
dertal linguistic capacities and its consequences. Frontiers in Psychology 4:397. DOI: 10.3389/fpsyg
.2013.00397.
Givón, Talmy. 2002. The visual information-processing system as an evolutionary precursor to human lan-
guage. The evolution of language out of pre-language (Typological studies in language 53), ed. by
Talmy Givón and Bertram F. Malle, 3–50. Amsterdam: John Benjamins.
Harris, Eugene E. 2015. Ancestors in our genome: The new science of human evolution. Oxford: Oxford
University Press.
Kos, Miriam; Danielle Van Den Brink; Tineke M. Snijders; Mark Rijpkema; Barbara Franke;
Guillen Fernandez; and Peter Hagoort. 2012. CNTNAP2 and language processing in healthy in-
dividuals as measured with ERPs. PLoS ONE 7.10:e46995. DOI: 10.1371/journal.pone.0046995.
Lewontin, Richard C. 1998. The evolution of cognition: Questions we will never answer. An invitation to
cognitive science: Methods, models, and conceptual issues, vol. 4, ed. by Don Scarborough and Saul
Sternberg, 107–32. Cambridge, MA: MIT Press.
Progovac, Ljiljana. 2015. Evolutionary syntax. (Oxford studies in the evolution of language.) Oxford: Ox-
ford University Press.

Linguistics Program
Wayne State University
5057 Woodward
Detroit, MI 48202
[[email protected]]