Language as symbolic behaviour and languages as symbolic systems left no traces in the archaeological record of hominin and early human societies – nor could it be otherwise. The most problematic aspect of research on human cognitive evolution is precisely that whether or not our ancestors possessed a given cognitive capacity (e.g., language) must be inferred indirectly from a sparse, heterogeneous set of datapoints, spanning several continents and millions of years. The challenge of interpreting and integrating these data, against the backdrop of linguistics, neuroscience, genetics, and primate ethology, has been taken up by many in recent years, often with only moderate success. The new book by Ronald J. Planer and Kim Sterelny, From signal to symbol: The evolution of language, is a clear step forward for research in this tradition. It provides a plausible, empirically grounded account of language evolution in a captivating format, striking a difficult balance between brevity, clarity, and richness of documentation and detail.
The book opens with some considerations on the requirements of theories of language evolution – that they be incremental and embedded in accounts of hominin and human social, technological, and economic evolution – and on the inference patterns that those theories rely upon, such as “(…) to use the material record to identify the availability of cognitive and social resources for particular communicative capacities [and] to identify communicative needs that select for those capacities. If capacities were both available and useful, they [hominins and humans] probably had them” (p. 27). Chapter 1 briefly presents the evolutionary story favoured by Planer and Sterelny (P&S), anticipating many ideas and themes developed in the book. Chapters 2–7 describe the major evolutionary steps and transitions from ‘the Great Ape baseline’ to (gestural) protolanguages, from composite signs to grammatical structure, and from sign to speech.
In Chapter 2, P&S discuss research on vocal and gestural communication in great apes and monkeys, and conclude that the expansion of signalling skills that eventually led to human language likely began in the gestural modality. The genetic and neurobiological infrastructure that supports language appears to be largely shared by modern humans with early hominins and even with great apes. However, this does not imply shared cognitive phenotypes or behaviours: cultural evolution, in the context of social, technological, and economic transitions, is needed to realise (latent) cognitive capacities in specific forms, and to explain how language-ready hominins became language-equipped humans. The authors accept that adaptive phenotypic plasticity played a key role in the evolution of language: if a behavioural innovation is successful, it is initially stabilised via learning; if it is persistent and useful enough, selection will favour genetic variants that allow more rapid and reliable acquisition of the behaviour, or that expand it or find new uses for it (p. 27; West-Eberhard, Reference West-Eberhard2003). The ‘language mosaic’ – the suite of cognitive-behavioural capacities that support language – is firmly rooted in primate and hominin biology. It is changes to the lifeways of our ancestors that should be invoked to shed new light on the evolutionary trajectory that led from great ape gestures to modern language. Simplifying somewhat the authors’ narrative, three major shifts were involved.
The first is the transition from simple to composite signs (Chapter 4). The evolution of bipedalism, which was likely completed at the end of the Pliocene, 2.5 million years ago, projected Homo erectus in open environments, where cooperative planning was crucial to survival. Bipedalism freed the upper limbs and selected for new hand movements and grips with greater power and precision. The Erectines had cognitive and motor abilities sufficient for a moderately sophisticated gestural protolanguage, and their lives would benefit from a communication system with structured signals and meanings, displaced reference, and an open repertoire of basic signs, to which new ones could be added. At this stage, much of the ‘language mosaic’ was in place and in use: early protolanguages required short-term and semantic memory, executive functions, social learning, and on-line processing, such as the capacity to integrate information across sensory modalities and over time. Other sophisticated cognitive skills are entailed by the lifeways of erectine hominins, who practiced ambush hunting (requiring information sharing, coordination, specialised weapons, self-control, and social tolerance), central-place foraging, division of labour and food sharing, the making of safe sites, collective vigilance and defence, and toolmaking. Against this background, P&S suggest that early composite signs emerged as combinations of a point and an iconic gesture. If A requests some object by using an iconic gesture and B misunderstands, then A must cope with communicative failure, for example, by producing a point to the desired object. In further interactions, A might try to prevent miscommunication by adding the point immediately after the iconic sign, in one fluent gesture. From this simple composition of signs, further expansions may have followed: for example, the addition of a third sign and the replacement of the point with a second iconic gesture. (Symbols, or arbitrary signs, emerged only considerably later, according to the authors, with the transition to the vocal-auditory modality.)
On this view, early protolanguages had linear grammars: signs were concatenated one after the other, and the (linear) distribution of tokens determined their relationships in interpretation. Chapter 5 is dedicated to the next transition, from linear to hierarchical grammars. The evolution of toolmaking brought complex, long, and precise sequences of manual actions under executive control, along with the capacity to predict the effects of such actions during toolmaking. The cognitive capacities for syntactic representation and processing, according to P&S, build on pre-existing hierarchical processing skills in apes, which were further expanded for stone toolmaking in our lineage. The abilities to parse sequences of sounds or acts into parts, and to interpret the whole as a function of the parts and of the order in which they occur (‘proto-compositionality’) released early humans from ‘semantic soup’ strategies for the interpretation of complex signals, where “(…) merely activating representations corresponding to the various words or phrases in the sentence is sufficient to understand its meaning when there is only one plausible way of assembling those representations into a coherent scenario” (p. 123). P&S argue that it was only when “a semantic soup strategy became unreliable that syntactic and/or other grammatical devices emerged” (p. 127). Changes in the social and economic conditions of early human communities selected for the extension of hierarchical processing skills from stone toolmaking to communication. With the expansion of social, technological, and economic complexity came the need to overtly mark time, space, agency, and other relational aspects of transactions or events. This recruited hierarchical processing skills for both production and comprehension, and selected for sign systems in which cues to hierarchical structure are also overtly marked. (I return to the hypothesised connection between complex actions and hierarchical syntax below.)
The third major transition, according to P&S, transformed early signed protolanguages into multimodal protolanguages with a prominent vocal and auditory channel (Chapter 6). This shift may have started before hierarchical structure emerged. The authors link it to another revolution in human life: the domestication of fire and its implications for diet, social organisation, and time management. There is early evidence of human-made fire between 0.5 and 1 million years ago, and the neural and vocal infrastructures for speech and for auditory speech processing were likely available to the ancestors of modern humans and Neanderthals in the mid-Pleistocene. So, why did modern language apparently not emerge until about 200 thousand years ago? P&S note that anatomical structures ideal for chewing and swallowing tough foods are not well suited for controlling speech, and vice versa. However, that trade-off is less severe if foods are processed. Moreover, some forms of food preparation, such as cooking, imply the control of fire, which changes the social environment in ways that favoured communication. If processed or cooked foods reduce feeding time, then the mouth becomes more available for other activities (e.g., singing, laughing, and eventually talking), and communication becomes possible even when the hands are occupied. Dunbar & Gowlett (Reference Dunbar, Gowlett, Dunbar, Gamble and Gowlett2014) have argued that language originated by the fireside. According to P&S, protolanguages were already available in gestural form, but the firelight niche made the transition to spoken protolanguage possible.
There is much to like about From signal to symbol. It is a rich and detailed account of many facets of language evolution. It is truly incremental and truly embedded in some of the best research on primate, hominin, and human evolution. In this sense, the book succeeds in meeting the standards set by its authors. It is, moreover, epistemologically innovative and coherent, particularly in its use of specific inference patterns from data indicating that early humans had the cognitive capacities entailed by X (protolanguage, hierarchical structure, etc.) and that their environments selected for X, to the conclusion that they probably had X. The emphasis on human cooperation, and specifically on the evolution of obligate mutualistic cooperation and later of delayed-returns cooperation, as drivers of communicative and linguistic evolution is also plausible and very appealing. The book may not fully address fundamental linguistic questions, but that is also where most other accounts of language evolution fall flat. For example, P&S note that language is simultaneously a social phenomenon (i.e., a feature of communities of speakers, which individual learners join) and an individual trait (a capacity for learning and processing one or more languages): “Evolutionary theories of language sometimes take the social phenomenon as their primary target, sometimes the individual trait. An adequate theory needs to give an account of both, and of their interaction.” (p. 2). This is a tall order, and P&S’s book hints, often with great clarity and depth of insight, at possible solutions. However, the emerging narrative tends to presuppose, rather than explain, cognitive phenotypes, which are often viewed as having been repurposed ‘for’ (proto)language, and little is said about how properties of natural languages (e.g., syntactic categories, compositionality, and pragmatic principles) may have evolved.
The book presents a critique of neo-Peircean index-icon-symbol models (Everett, Reference Everett2017) of language evolution, on the grounds that indices are too diverse to form a sufficiently homogeneous class and serve as a springboard for protolanguage, and that icons can be highly abstract and require no fewer cognitive resources than symbols. P&S use instead signalling games as an evolutionary model of coordination and communication. This is a reasonable starting point for theorising about primate vocal calls, protolanguage, and human language, but P&S do not engage with the extensive theoretical, computational, and experimental literatures connecting signalling games to linguistics, in particular to lexical, referential, and compositional semantics (see Baggio, Reference Baggio2018; Moreno & Baggio, Reference Moreno and Baggio2015; Nowak & Baggio, Reference Nowak and Baggio2016). There is much potential for applications of signalling games to test the formal and empirical plausibility of the evolutionary trajectory delineated by P&S: for example, under what conditions on the complexity of the meanings to be expressed can signalling systems emerge featuring hierarchically structured signals? One should hope that the authors, or others inspired by them, will take up such challenge in future work.
Another area where P&S could have dug deeper is the connection between planning or complex actions and hierarchical syntax. They argue that “(…) great apes seem to have the basic capacities they need for simple syntax. In production, they can recall gestures from memory, choose distinct gestures, and launch sequences of them. As interpreters, (…) they can learn to recognize and use that sequential information” (p. 126), and that “(…) neither the evolution of our capacity to use linear syntax, nor the actual creation of linear syntax in our line, pose deep mysteries. The great apes have the capacity, or almost have it. (…) Where there seems to be a large gap between us and other animals, including our close primate relatives, is in the capacity to use hierarchical structure in communication” (p. 129). Yet grammars are defined over syntactic types (categories of words, such as N, V, and P, of phrases resulting from their combination, such as NP and VP, and, at least in some frameworks, of abstract elements, such as tense, and even silent elements, which are not expressed phonologically but may be interpreted semantically). P&S are justified when they question the assumption that because hierarchical structure is what formal grammars assign to strings, then speakers and hearers must internally represent that hierarchical structure when they process the same sentences. However, one of the reasons why such assumption is questionable is a lack of clear-cut evidence for the internal representation of syntactic types to begin with. Even though syntactic priming studies, for example, suggest that abstract syntactic structure is ‘psychologically real’, it has proved difficult even to find conclusive evidence for the distinction between nouns and verbs as grammatical classes in the brain (Vigliocco et al., Reference Vigliocco, Vinson, Druks, Barber and Cappa2011). Syntax is not only a matter of (hierarchically) structured sequences: it is the elements of those sequences (syntactic types) that are special, in that they (1) co-determine compositional meanings via specific interpretive mappings to semantic types (roughly, the elements of cognitive ontologies) and (2) introduce specific dependencies and constraints that rule out some (hierarchical) structures as ungrammatical (Moro, Reference Moro2014). We know little about whether and how syntactic types are represented in the brain and how they could be constructed during language learning. Theories of semantic interpretation that partly dispense with syntactic types may be conceivable, and could even provide a foundation for theories of language evolution. However, given our best theories in linguistics, the emergence of cognitive capacities for the representation, acquisition, and processing of syntactic types remains a missing piece in the evolutionary puzzle.
From signal to symbol succeeds in delivering an empirically motivated, realistic, and compelling account of the evolutionary origins of human language. What is more, P&S present the problems in their full complexity (Chapter 1 is titled An Intractable Challenge?), rejecting saltational and reductionist accounts. Necessarily, the book only scratches the surface of some of the hardest linguistic issues, like syntax and compositionality. However, to leave the reader wanting more is not at all a flaw: it is a virtue and an indication that the direction charted by P&S might be correct. The book is written in an informal yet precise style. The sources and materials are carefully selected, clearly presented, and seamlessly integrated. This is not a technical, specialist monograph, rather an accessible synthesis that offers enough ‘big picture’ to satisfy most generalist cognitive science readers, and enough details to keep connoisseurs turning (and annotating) the pages.
