The uniqueness and complexity of human language unavoidably demands an evolutionary explanation, which, after an eclipse of more than a century, has reappeared as a robust subject for scientific enquiry, especially since Pinker & Bloom (Reference Pinker and Bloom1990).Footnote ^[1] The host of proposals and scenarios roughly falls into either incremental theory, a Darwinian approach to language evolution through gradual stages, or a saltationist theory, a great leap resulting in the unique property of language. In contrast, in her book Cognition and Communication in the Evolution of Language Anne Reboul proposes a two-step approach to language evolution. According to the model, the most primitive part, the conceptual apparatus, which consists of concepts and core cognition, evolved first as a result of minor modification to the modern human perceptual system, and the resulting conceptual explosion leads to syntactical Merge, a recursive operation of set-formation with two objects into a single new object in line with the Minimalist Program, as an auto-organization process for Language of Thought. This, under social selection pressure, is in turn externalized for communication. This two-step approach to language evolution necessarily requires a dual account for language evolution. Drawing on insights from cognitive science, anthropology, comparative psychology and theoretical linguistics, the book is both inspirational and provocative to anyone interested in language evolution. It not only links human thought and language in the process of evolution, but also provides a starting point for tackling the incompatibility of incremental theory and saltationist theory, trying to dissolve the dilemma of language evolution.

The book is well-organized. After setting the criteria for language evolution theory in Chapter 1, the author discusses the necessity of a dual account for language evolution in Chapter 2, describing the specificity of human conceptual apparatus in Chapter 3, discussing the first and the second step of language evolution in Chapter 4, exploring the reasons of externalization in Chapter 5, and concluding in Chapter 6.

Chapter 1 sets out the neutral criteria for a theory of language evolution, whether it is the result of cultural evolution or biological evolution, or both. Language evolution is unlikely to be either a saltational or an incremental process, because short language evolution time makes gradual process implausible, and genomic proximity between humans and chimpanzees makes massive biological innovation also implausible. The present function of language as communication may be an unreliable guide to the primordial evolutionary motivations. Any discontinuity theory which sees human language as different from all other animal communication systems is incompatible with Darwin’s natural selection theory. However, Reboul holds that a gradual theory does not necessarily preclude discontinuity (otherwise there would be no innovation in biology). Discontinuities, on the contrary, are the goals to be explained within natural selection theory. As for language, it might be quantitatively discontinuous like thought, indicating that language may have gradually evolved from previous cognitive systems through natural selection, but qualitatively discontinuous like communication, indicating language is distinct from non-human animals’ communication. Reboul therefore distinguishes language as a communication system in the strong sense, where language evolved for communication, from language as a communication system in the weak sense, where language evolved first for thought and then exapted for communication. The latter approach, which, accordingly, would be perfectly compatible with natural selection, can provide a reasonable account of language evolution.

In Chapter 2, ‘The need for a dual account of language evolution’, Reboul argues that the core combination of discrete infinity (infinite number of sentences), semanticity (sentences with different contents), and decoupling (language independent of situation), which conjointly makes language unique among animal communicational systems, is central to any account of language evolution theory. Signals in animal communication systems, due to lacking compositionality, are holistic and perennial in nature. Two models of language in the strong sense, which ignore the compositionality of language and thus face some counterarguments, are discussed. The code model of communication, largely borrowed from information theory, assumes that language is an encoding–decoding system, actually presupposing that the message pre-exists its encoding and decoding. Any theory of language evolution of this kind actually targets external languages, which are clearly the products of cultural but not biological evolution. Given linguistic creativity, the model meets problems such as perenniality of signals and ambiguity in pragmatic inference. Moreover, the model may suggest that language acquisition is a general process like other animal communication systems, which are strongly phylogenetic and hence not acquired. On the other hand, ostensive theory, heavily based on Relevance Theory (Sperber & Wilson Reference Sperber and Wilson1995), sees language as an ostensive-inferential communication system, which is actually the result of double communicative and informative intention. The model also meets problems such as pragmatic stability and illocutionary force in linguistic communication.

In Chapter 3, ‘The specificity of the human conceptual apparatus’, Reboul argues that human cognition, with the change in the perceptual preference towards a global treatment of information rather than at a local level, is significantly different from that of non-human animals in social abilities and general cognitive abilities, notably in the ability for abstraction. In particular, non-human animals, due to their poor ability to conceptualize and categorize objects, might only discriminate relevant objects. In contrast, humans as a generalist species, using their physical and cognitive abilities to survive in widely different environments, can categorize the same objects at different levels, forming extensive collections of concepts and rich conceptual hierarchies at basic, subordinate, and superordinate levels, far beyond the direct experience and biological constraints. This gives humans access to a large number of concepts, which, in Reboul’s view, are the units to be combined in human thought, characterized by conceptual stability (systematicity) and semantic compositionality. Concepts (at least atomic concepts such as BROWN or COW versus BROWN COW) cannot be stores of information, nor can they depend on subjective experience. The contents of a concept must be interpretable as the results of compositionality, making it highly unlikely that concepts are prototypes, exemplars or theories, though concepts are involved in some cognitive processes. Concept possession is the ability to think thoughts involving concepts. As in semantics, concepts should be connected. While an intension view of connecting concepts is something like a definition, which is epistemologically unlikely, an extension view ensures both systematicity and compositionality of concepts. This leads to a conceptual explosion and triggers a syntactic process, Merge $\unicode[STIX]{x1D6FC}$ , allowing humans to link concepts together in thought and resulting in the unique ability for strong decoupling, the cognitive process not entertained as a means to any physical action.

In Chapter 4, ‘Merge and the lexicalization of concepts’, Reboul argues that Merge $\unicode[STIX]{x1D6FC}$ as a self-organizing process is applied in both externalization processes. Adopting the exo-skeletal approach, Reboul argues that syntactic objects offer poor information except reference, and that their interpretations are attributed to the syntactic position they occupy. The first step in language evolution is from conceptual apparatus to Language of Thought. The purely referential and atomic concepts of human being are organized through Merge $\unicode[STIX]{x1D6FC}$ , whose result is transferred to the interpretation system, the semantic component (SEM). At the same time, core cognition such as folk-biology and folk-psychology, which strongly condition the perception of reality, may play a sub-categorizing role and allow for finer-grained thought during transfer. The second step in language evolution is the externalization from Language of Thought into language as a communication system, with the addition of the phonological component (PHON) at the externalization. Reboul transforms the question of externalization into the question of how tight a fit there exists between concepts and lexical items. Language and even perception, according to Sapir–Whorf Hypothesis, are strongly constrained by language. Through a careful analysis of various experiments, Reboul argues that primitive concepts are independent of language. Furthermore, cognitive neuroscience experiments show that the localization of categorical perception in infants’ right hemisphere delocalizes to adults’ left hemisphere after acquisition of the relevant lexicon, suggesting that categorical perception is independent from language too. Reboul further argues that the innovation during externalizations is the emergence of the functional lexicon items, which link core cognition which conditions the perception of reality, such as classifiers with the countability of nouns.

Chapter 5, ‘A mildly Machiavellian view of communication and the argumentative theory of reason’, deals with reasons for externalization from Language of Thought to language as communication. Reboul rejects the altruism hypothesis in language evolution (Tomasello Reference Tomasello2008). Linguistic communication, though indeed cooperative in the Gricean sense, is basically mutualistic, in the sense that linguistic communication is beneficial to all participants. Information is a precious commodity, and information sharing would be detrimental to the speaker, which is implausible from an evolutionary point of view. This is evidenced by two cross-linguistically ubiquitous instances of implicit communication, conversational implicature and presupposition, both unpredicted by Universal Grammar or general mechanisms of linguistic communication. Compared with explicit communication, implicit communication is apparently costly and inefficient for both speaker and hearer. Reboul argues that it is mildly Machiavellian or manipulative, aiming at inducing in the hearer a behavior that is beneficial to the speaker. The externalization of Language of Thought, therefore, is due to changes in social organization in modern humans. In the ‘hunter-gatherer’ society, for example, language of thought was externalized for collective decision-making processes, allowing individuals to formulate arguments and to give reasons as to why their own opinions should prevail over opposite opinions. Here the notion of argumentation comes in. The Argumentative Theory of Reasoning, based on the idea that human reasoning is social and communicative in nature, thus links reasoning and (external) language, supporting the idea that Language of Thought is externalized for manipulation.

Chapter 6 concludes with a summary, followed by some robust suggestions for interactionist approaches to language evolution, since language is far from being a monolithic ability, but rather the result of pre-existing abilities and innovative Merge. Finally, Reboul briefly traces back the evolutionary history of the Homo lineage, showing that the two-step approach is consistent with this evolutionary history.

In sum, Reboul offers an admirably persuasive account of the two-step approach to language evolution, making explicit the connections between language and human thought. Chomsky, who very often states notions such as ‘language of thought’ or ‘meaning with sounds’ (e.g. Berwick & Chomsky Reference Berwick and Chomsky2016), while focusing on the generation of language, does not go very far concerning the relation between language and thought. On the other hand, the syntactic operation of Merge, a crucial theoretical constituent in current biolinguistics, certainly requires objects to be merged. Otherwise, language evolution would be equal to the mutation of Merge. However, Chomsky pays little attention to the objects of Merge (see Boeckx Reference Boeckx2015), but sometimes employs vague terms such as ‘word-like’ (Berwick & Chomsky Reference Berwick and Chomsky2016). Crucially Reboul identifies the objects of Merge from an evolutionary perspective.

Methodologically, the interactionist approach to language evolution, taking both internal and environmental factors into consideration, offers the potential to explain the uniqueness of language without contradicting Darwin’s natural selection. It also provides a theoretical consideration to dissolving the contradictions between linguistic universals, the fundamental assumptions of Generative Grammar, and linguistic diversity, a focus such as in cognitive linguistics. In the two-step model of language evolution, Merge and strong decoupling, which are the results of natural evolution, are biologically unique to human beings, while most cross-linguistic variations, due to derivational dynamics and grammaticalization, and variable extensions of notion derived from core cognition systems, may occur during the transfers to the interfaces.

Finally, a discrete system such as human language requires a recursive algorithm, which makes Merge a mechanical procedure, characterizing the infinite generative capacity of language. If a conceptual explosion is on the right track, recursion, which is argued to have evolved recently (Hauser, Chomsky & Fitch Reference Hauser, Chomsky and Tecumseh Fitch2002), may appear at this stage. The book would thus make for a more satisfying read if some discussion of recursion were included. This aside, the book offers very thought-provoking insights into a theory of language evolution.