In his latest book, Michael A. Arbib argues that human language evolved from animal communication with the help of ‘complex imitation’. A basic form of gestural communication was transformed into vocal communication and from there, cultural evolution continued to evolve basic protolanguages into the kind of complex languages that we have today. The argument relies heavily on how mirror neurons in the brain make it possible for monkeys, apes and humans to recognize the manual actions of others.

The book comprises 13 chapters. The first five make up Part 1, ‘Setting the stage’, and the following eight make up Part 2, ‘Developing the hypothesis’.

In chapter one, Arbib introduces schema theory, a model of how brain functions (human and animal) can be modelled as action–perception schemas. Arbib shows how this model can be applied to animal behaviour. After that, vision and dexterity are treated in detail in the model.

In the second chapter the reader is introduced to Arbib's perspectives on human languages. A large part of the chapter is devoted to the importance of semantic compositionality and the relevance of an autonomous syntax. Chomsky's approach to the structure of language is presented. However, basing his arguments on the action–perception schemas introduced in chapter one, Arbib considers construction grammar to be the most suitable framework for linguistic description.

Chapter three presents studies on communication in groups of monkeys and apes. The focus is on vocalization and gestures. Arbib shows, on the one hand, how the small repertoire of monkeys’ vocalization is far from what can be called a language; apes, on the other hand, have communicative manual gestures. In addition, these gestures vary not only between species, but also between groups within the same species. This is, according to Arbib, a clear indication that these gestures are not innate, but instead ‘invented’ and transmitted via learning to other members of the group. Arbib reviews different hypotheses of how these gestures arise. However, the set of gestures that apes use is very limited and, crucially, Arbib points out, nothing in the gestural system suggests that there is any kind of syntax or compositionality. Arbib also mentions some studies where language has been taught to apes. Again, his conclusion is that their repertoire is extremely limited, though larger than the one shown by apes in the wild, and that there is no sign of either compositionality or syntax. In fact, Arbib's claim is that even though these apes have been trained ‘to use “language” they are not really using language since their vocabulary is limited and they have no sense of syntax’ (p. 150).

In chapter four Arbib outlines the differences and similarities between monkey brains and human brains. The important similarity is the fact that the centre for grasping in the monkey brain corresponds approximately to Broca's area in the human brain. However, the corresponding area (F5) in the human brain is vastly more developed and includes some other regions, lacking in the monkey brain. The key point for Arbib is that this is an area in which there is a mirror neuron system. Mirror neurons were discovered in the late eighties in the macaque monkey brain. Mirror neurons are neurons that fire both when the monkey performs an action, such as grasping an object, and when the monkey watches another monkey, or a human, perform the same action.

Chapter five deepens the description of the mirror neuron system. Arbib reviews the studies on mirror neurons in the human brain. Since it is impossible to look at separate neurons without neurosurgery in studies of humans, the ways of measuring firing mirror neurons in the human brain are not as accurate as the way in which the mirror neurons have been studied in monkeys. As a consequence, our knowledge of the human mirror neuron system is limited. There are studies showing that mirror neurons fire not only by watching someone grasping, but also in relation to orofacial (mouth and face) expressions and the sounds of actions (such as peanut breaking). Arbib presents a hypothesis of how mirror neurons play a central role in the actions of both the self, and also in the observation of others. Since mirror neurons are intimately connected to the motor system (as explained in detail in chapter four), they provide humans with a way of understanding actions of others. This chapter presents the gist of Arbib's hypothesis, that mirror neurons are central in what is called ‘complex imitation’ and that complex imitation in turn can transform simple gestures of the kind apes have, into a protolanguage. This chapter concludes the first part of the book.

The second part starts with chapter six, which is partly a summary of the previous chapters, partly an introduction to the cultural and biological evolution of Homo sapiens, and partly a summary of the chapters that lie ahead. In this chapter Arbib presents his mirror system hypothesis in detail. Arbib assumes that the mechanisms that support language in the brain are evolved on top of a basic mechanism that was not originally related to language (the mirror system in macaques). This mirror system made it possible to generate and recognize actions and it provides the evolutionary basis for language parity – the property that an utterance (or gesture) means more or less the same thing for both speaker (signer) and hearer (viewer). The hypothesis is as follows (p. 174). The pre-hominid brain was equipped with two components: (i) a mirror system for grasping – a system that existed in a common ancestor of humans and monkeys, and (ii) a simple imitation system for grasping. This system was present in the common ancestor for apes and humans. The evolution from pre-hominid to hominid brought about another two components. The first component is a complex imitation system. According to Arbib complex imitation involves, firstly, combining the ability to recognize some one else's behaviour as a set of familiar movements with the ability to use this recognition to repeat the behaviour and ‘(more generally) to recognize that another's performance combines actions that can be imitated at least crudely by variants of actions already in the repertoire, with increasing practice yielding increasing skill’ (ibid.). This newly developed imitation system paved the way for what Arbib calls Proto-signs. According to his hypothesis, hominids used reduced form of actions, ‘pantomime’ to communicate. The imitation system reduced the pantomime to conventionalized signs, proto-signs. The second component is protosigns. Protosign is an open ended manual-based communication system. According to Arbib, the mirror neuron system intimately connects manual dexterity and language in Broca's area. Eventually the mirror system of the manual system took control, or ‘invaded’ in Arbib's term, the vocal apparatus. The result is protospeech and multimodal protolanguage. The manual system and the speech system feed each other in an ever growing ‘spiral’. These four components make the human brain ‘language-ready’ and the final step into ‘modern’ languages is not a biological evolution according to Arbib. The transition from naming of actions and objects into compositional semantics and syntax is the effect of a cultural evolution in Homo sapiens.

Chapter seven provides the details of Arbib's hypothesis of complex imitation. According to Arbib, complex imitation is one of the most important capabilities in distinguishing humans from other primates. The distinctions between simple and complex imitation is the possibility of not only acquiring an observed plan of action by a long process of watching and trying again and again (this is what apes are capable of), but also the ability to just observe a behaviour and acquire the plan, without performing the action. This possibility of acquiring a plan without actually performing the action itself, Arbib calls complex imitation.

Chapter eight outlines Arbib's hypothesis of how the ability of complex imitation over time developed into pantomime. The gesture system comprised reduced forms of actions to convey aspects of other actions, objects, emotions or feelings. Arbib provides the following example: ‘Flapping one's arms in a “wing-like” manner could indicate an object (a bird), or an action (flying), or both (a flying bird)’. Over time, and with use, the pantomimes were conventionalized and the result was a system of protosigns.

Chapter nine develops Arbib's hypothesis of how protosigns evolved into protospeech. Arbib assumes that protosigns played a crucial role in the evolution of protospeech. Over time, and with use, the manual gestures came to be accompanied by vocalizations. Arbib's hypothesis is based on the fact that language is not limited to vocalization; the signed languages of deaf people are equally important to take into account. According to the hypothesis, the increasingly spoken protovocabulary might have provided evolutionary pressure and lead to the evolution of the vocal apparatus and the neural system that controls the rapid production of phonemes.

Chapter ten builds on Wray's (1998, 2000) views on how the ‘holophrases’ of protosign/speech, were gradually split up, via ‘fractionation’, into morpheme-like parts. Holophrases are words with no internal structure but with a complex semantics. According to Arbib a holophrase such as ‘koomzash’ might mean ‘take your spear and go around the other side of that animal, and we will have a better chance together of being able to kill it’ (p. 258). Eventually, there will be strings in holophrases that can be tied to specific objects/items, at first by pure coincidence, and later as an active strategy. The number of ‘words’ will increase for each generation, and the growing vocabulary introduces the need for a syntax that can combine the different words into new meanings.

Chapter eleven is an expose of how children acquire language. Arbib frames the language evolution of two-year-old children in a construction grammar. He claims that even though children learn single words, such as ‘milk’, they treat them as holophrases. In fact, ‘milk’ has the meaning of ‘want milk’, whose pieces have no separate meaning (p. 287).

Chapter twelve discusses the emergence of new languages. Arbib focusses on two sign languages, Nicaraguan Sign Language (NSL) and Al-Sayyid Bedouin Sign Language (ABSL), neither of which developed as variants of existing sign languages. Arbib rightfully points out that neither of these two languages evolved without the influence of speakers of other languages. In the case of NSL, the Nicaraguan signers were influenced by teachers from various linguistic backgrounds. In the case of ABSL, some of the signers are hearing people with a fully developed spoken language, Arabic. In addition, the crucial difference between these new languages and the language that evolved through cultural evolution in the history of the human species is that the new languages all develop in a context in which it is obvious to the speakers that there is such a thing as language. This is quite a different thing from developing a language when there is nothing in the social environment that indicates that such a thing can exist. In this chapter, the development of creole languages from pidgin languages is also briefly discussed.

Chapter thirteen is the final chapter and here Arbib presents archaeological findings, such as tool making techniques and engravings, and ties them to the development of the mirror neuron system. He also presents Heine & Kuteva's (2007) hypothesis of the historical development of syntactic categories. Heine & Kuteva assume that all grammatical markers have a common origin in the category noun. According to Arbib, this process was active already when protolanguage developed into language and has been ever since.

Arbib's book is well written and engaging. Especially the chapters dealing with mirror neurons and how they function are very interesting. My major objection to Arbib's hypothesis of how language evolved is that it hinges on too many weak assumptions. The crucial point is the mirror system which takes a communication system based on gestures into a vocal system. As Arbib himself says in the book, the fact that there is a mirror system near Broca's area may be a coincidence. We do not even know if there are other mirror systems in the human brain. It is also difficult to imagine what a communication system based on pantomime would be able to communicate. It seems that the most obvious way of using gestures would be by pointing. Even though pointing is a system limited to the ‘here and now’ it is difficult to see how flapping one's arms would go beyond the here and now. In addition, how would one communicator be able to know that a receiver has even remotely understood the message? Burling's (2007) hypothesis of how pointing accompanied by vocal sounds may have evolved into protolanguage seems much more plausible. In fact, there is nothing in the mirror neuron hypothesis that contradicts Burling's hypothesis, except the large emphasis on pantomime, but Arbib does not even mention Burling's work. Another criticism is the lack of linguistic argumentation. Arbib stresses the importance of compositionality and syntax, and even claims that this is what separates human language from apes’ communication. Even so, he attributes the emergence of syntax and compositionality to cultural evolution, even though he hypothesizes that the split between ape and human communication comes long before the advent of syntax. Also, Arbib's adoption of construction grammar as a useful framework seems somewhat ad hoc. In most linguistic aspects, the argumentation in the research he reviews is not compatible with a construction grammar approach. It is also not clear why there should be a stage of holophrases. In the review of communication among apes, there does not seem to be any indication of holophrases. And once they are in place, it is unclear why speakers will start separating them. As Bloom (2000) points out in relation to the famous ‘gavagai’ problem, prelinguistic infants are strongly biased to parse the world into discrete bounded entities, i.e. not holophrases, and that these entities, ‘Spelke objects’ (Spelke 1994) correspond to the notion of ‘object’ used in theories of language acquisition. From a linguistic point of view, language in Arbib's sense is a matter of lexical semantics, and the syntax that he discusses never goes beyond mere argument-predicate structure. All the semantic intricacies of modality, negation, quantification, etc. are not mentioned at all. In a full-fledged theory of language evolution, such aspects of language should be taken into consideration. Also, it is not immediately evident how these notions can be accounted for in the kind of embodied perspective Arbib has. What reflexes in vision and motion do these abstract notions have? And even if syntax and compositionality are indeed the result of cultural evolution, Arbib does not spell out the details of such an evolution so the reader can judge whether it is credible or not.

Readers familiar with Arbib's work will not find anything new in this book. Most of the chapters are based on already published material. This is probably why there is no real connection between them. Even if one does not agree with Arbib on how the human brain got language, the book may interest the intended audience, which according to the publisher is educated lay readers, as well as researchers in linguistics, cognitive neuroscience, neurolinguistics, natural language processing, primatology, and anthropology. However, for a linguist like myself, the most intriguing aspect of the book is the one dealing with the human brain and the mirror neuron system.