Language Evolution and Computation Bibliography

This paper proposes a coevolutionary scenario on the origins of compositionality and word order regularity in human language, and illustrates it using a multi-agent, behavioral model. The model traces a bottom-up process of syntactic development; artificial agents, by iterating local orders among lexical items, gradually build up basic constituent word order(s) in sentences. These results show that structural features of language (e.g. syntactic categories and word orders) could have coevolved with lexical items, as a consequence of general learning mechanisms (e.g. pattern extraction and sequential learning) initially not language-specific.

Arbitrariness and systematicity are two of languageas most fascinating properties. Although both are characterizations of the mappings between signals and meanings, their emergence and evolution in communication systems has generally been explored independently. We present an experiment in which both arbitrariness and systematicity are probed. Participants invent signs from scratch to refer to a set of items that share salient semantic features. Through interaction, the systematic re-use of arbitrary signal elements emerges.

This paper contrasts two influential theoretical accounts of language change and evolution a Iterated Learning and Social Coordination. The contrast is based on an experiment that compares drawings produced with Garrod et alas (2007) apictionarya task with those produced in an Iterated Learning version of the same task. The main finding is that Iterated Learning does not lead to the systematic simplification and increased symbolicity of graphical signs produced in the standard interactive version of the task. A second finding is that Iterated Learning leads to less conceptual and structural alignment between participants than observed for those in the interactive condition. The paper concludes with a comparison of the two accounts in relation to how each promotes signs that are effi cient, systematic and learnable.

Human communication is often thought about in terms of transmitted messages in a conventional code like a language. But communication requires a specialized interactive intelligence. Senders have to be able to perform recipient design, while receivers need to be able to do intention recognition, knowing that recipient design has taken place. To study this interactive intelligence in the lab, we developed a new task that taps directly into the underlying abilities to communicate in the absence of a conventional code. We show that subjects are remarkably successful communicators under these conditions, especially when senders get feedback from receivers. Signaling is accomplished by the manner in which an instrumental action is performed, such that instrumentally dysfunctional components of an action are used to convey communicative intentions. The findings have important implications for the nature of the human communicative infrastructure, and the task opens up a line of experimentation on human communication.

Communication is a fundamentally interactive phenomenon. Evolutionary biology recognises this fact in its definition of communication, in which signals are those actions that cause reactions, and where both action and reaction are designed for that reason. Where only one or the other is designed then the behaviours are classed as either cues or coercion. Since mutually dependent behaviours are unlikely to emerge simultaneously, the symmetry inherent in these definitions gives rise to a prediction that communication will only emerge if cues or coercive behaviours do so first. They will then be co-opted for communication. A range of case studies, from animal signalling, evolutionary robotics, comparative psychology, and evolutionary linguistics are used to test this prediction. The first three are found to be supportive. However in the Embodied Communication Game, a recent experimental approach to the emergence of communication between adult humans, communication emerges even when cues or coerced behaviours are not possible. This suggests that humans are exceptional in this regard. It is argued that the reason for this is the degree to which we are able and compelled to read and interpret the behaviour of others in intentional terms.

Although rapidity of fading has been long identified as one of the crucial design features of language, little is known about its effects on the design of communication systems. To investigate such effects, we performed an experiment in which pairs of participants developed novel communication systems using media that had different degrees of rapidity of fading. The results of the experiment suggest that rapidity of fading does not affect the pace with which communication systems emerge or the communicative effi cacy of the emerged systems. However, rapidity of fading seems to affect the design of these systems. In particular, communication systems implemented in the more rapidly fading medium exhibited a higher degree of combinatorial reuse of their forms than systems implemented in the medium that faded more slowly. These results suggest that the design of language might be constrained by subtle relations the presence of which can be ascertained only through direct experimental manipulation. Human communication systems crafted today in the laboratory can provide new insights into the design of natural languages.

Recent work on the emergence and evolution of human communication has focused on getting novel systems to evolve from scratch in the laboratory. Many of these studies have adopted an interactive construction approach, whereby pairs of participants repeatedly interact with one another to gradually develop their own communication system whilst engaged in some shared task. This paper describes four recent studies that take a different approach, showing how adaptive structure can emerge purely as a result of cultural transmission through single chains of learners. By removing elements of interactive communication and focusing only on the way in which language is repeatedly acquired by learners, we hope to gain a better understanding of how useful structural properties of language could have emerged without being intentionally designed or innovated.

Computational simulations have provided evidence that the use of linguistic cues as group markers plays an important role in the development of linguistic diversity shortcite (Nettle & Dunbar, 1997; Nettle, 1999). Other simulations, however, have contradicted these findings (Livingstone & Fyfe, 1999; Livingstone, 2002). Similar disagreements exist in sociolinguistics (e.g. Labov, 1963, 2001; Trudgill, 2004, 2008a; Baxter et al., 2009). This paper describes an experimental study in which participants played an anonymous economic game using an instant-messenger-style program and an artificial aalien languagea. The competitiveness of the game and the frequency with which players interacted were manipulated. Given frequent enough interaction with team-mates, players were able to use linguistic cues to identify themselves. In the most competitive condition, this led to divergence in the language, which did not occur in other conditions. This suggests that both frequency of interaction and a pressure to use language to mark identity play a significant role in encouraging linguistic divergence over short periods, but that neither is suffi cient on its own.

It is widely assumed that language in some form or other originated by piggybacking on pre-existing learning mechanism not dedicated to language. Using evolutionary connectionist simulations, we explore the implications of such assumptions by determining the effect of constraints derived from an earlier evolved mechanism for sequential learning on the interaction between biological and linguistic adaptation across generations of language learners. Artificial neural networks were initially allowed to evolve ``biologically'' to improve their sequential learning abilities, after which language was introduced into the population. We compared the relative contribution of biological and linguistic adaptation by allowing both networks and language to change over time. The simulation results support two main conclusions: First, over generations, a consistent head-ordering emerged due to linguistic adaptation. This is consistent with previous studies suggesting that some apparently arbitrary aspects of linguistic structure may arise from cognitive constraints on sequential learning. Second, when networks were selected to maintain a good level of performance on the sequential learning task, language learnability is significantly improved by linguistic adaptation but not by biological adaptation. Indeed, the pressure toward maintaining a high level of sequential learning performance prevented biological assimilation of linguistic-specific knowledge from occurring.

In this paper, we propose two agent-based simulation models for the evolution of language in the framework of evolutionary language games. The theory of evolutionary language games arose from the union of evolutionary game theory, introduced by the English biologist John Maynard Smith, and language games, developed by the Austrian philosopher Ludwig Wittgenstein. The first model proposed is based on Martin Nowak's work and is designed to reproduce and verify (or refute) the results Nowak obtained in his simplest mathematical model. For the second model, we extend the previous one with the introduction of a world where the languages live and evolve, and which influences interactions among individuals. The main goal of this research is to present a model which shows how the presence of a topological structure influences the communication among individuals and contributes to the emergence of clusters of different languages.

The majority of the extant languages have one of three dominant basic word orders: SVO, SOV or VSO. Various hypotheses have been proposed to explain this word order bias, including the existence of a universal grammar, the learnability imposed by cognitive constraints, the descent of modern languages from an ancestral protolanguage, and the constraints from functional principles. We run simulations using a multi-agent computational model to study this bias. Following a local order approach, the model simulates individual language processing mechanisms in production and comprehension. The simulation results demonstrate that the semantic structures that a language encodes can constrain the global syntax, and that local syntax can help trigger bias towards the global order SOV/SVO (or VOS/OVS).

Social learning is a source of behaviour for many species, but few use it as extensively as they seemingly could. In this article, I attempt to clarify our understanding of why this might be. I discuss the potential computational properties of social learning, then examine the phenomenon in nature through creating a taxonomy of the representations that might underly it. This is achieved by first producing a simplified taxonomy of the established forms of social learning, then describing the primitive capacities necessary to support them, and finally considering which of these capacities we actually have evidence for. I then discuss theoretical limits on cultural evolution, which include having sufficient information transmitted to support robust representations capable of supporting variation for evolution, and the need for limiting the extent of social conformity to avoid ecological fragility. Finally, I show how these arguments can inform several key scientific questions, including the uniqueness of human culture, the long lifespans of cultural species, and the propensity of animals to seemingly have knowledge about a phenomenon well before they will act upon it.

If protolanguage was a holistic system where complex meanings were conveyed using unanalysed forms, there must be some process (analysis) which delivered up the elements of modern language from this system. This paper draws on evidence from computational modelling, developmental and historical linguistics and comparative psychology to evaluate the plausibility of the analysis process. While some of the criticisms levelled at analysis can be refuted using such evidence, several areas are highlighted where further evidence is required to decide key issues. More generally, the debate over the nature of protolanguage offers a framework for developing and showcasing a modern, evidence-based evolutionary linguistics.

Two opposing accounts of early language evolution, the compositional and the holistic, have become the subject of lively debate. It has been argued that an evolving compositional protolanguage would not be useful for communication until it reached a certain level of grammatical complexity. This paper offers a new, discourse-oriented perspective on the debate. It argues that discourse should be viewed, not as a level of language structure beyond the sentence , but as sequenced communicative behaviour, typically but not uniquely involving language. This provides for continuity from exchanges making use of simple communicative resources such as single words and gestures to those making use of complex grammatical conventions. Supporting evidence comes from child language and from original experiments with adults using constrained language systems. The paper shows that the utility of emerging compositional language is not dependent on some critical level of complexity, and so defends the compositional account.

We approach the issue of holophrasis versus compositionality in the emergence of protolanguage by analyzing the earliest combinatorial constructions in child, bonobo, and chimpanzee: messages consisting of one symbol combined with one gesture. Based on evidence from apes learning an interspecies visual communication system and children acquiring a first language, we conclude that the potential to combine two different kinds of semiotic element '' deictic and representational '' was fundamental to the protolanguage forming the foundation for the earliest human language. This is a form of compositionality, in that each communicative element stands for a single semantic element. The conclusion that human protolanguage was exclusively holophrastic '' containing a proposition in a single word '' emerges only if one considers the symbol alone, without taking into account the gesture as a second element comprising the total message.

A modular analysis of spontaneous language use provides support for the existence of an identifiable step in language evolution, protolanguage. Our suggestion is that a grammarless form of expression would have evolved to signal unexpected events, a behavior still prevalent in our species. Words could not be so specific as to refer to whole, non-recurring, situations. They referred to elements such as objects or locations, and the communicated event was inferred metonymically. Compositionality was achieved, without syntax, through multi-metonymy, as words referring to elements of the same situation were concatenated into proto-utterances.

The Complex-First Paradox regards the semantics of nouns and consists of a set of together incompatible, but individually well confirmed propositions about the evolution and development of language, the semantics of word classes and the cortical realization of word meaning. Theoretical and empirical considerations support the view that the concepts expressed by concrete nouns are more complex and their neural realizations more widely distributed in cortex than those expressed by other word classes. For a cortically implemented syntax ``semantics interface, the more widely distributed a concept s neural realization is, the more effort it takes to establish a link between the concept and its expression. If one assumes the principle that in ontogeny and phylogeny capabilities demanding more effort develop, respectively, evolve later than those demanding less effort, the empirical observation seems paradoxical that the meanings of concrete nouns, in ontogeny and phylogeny, are acquired earlier than those of other word classes.

This paper challenges recent assumptions that holophrastic utterances could be planned, processed, stored and retrieved from storage, focussing on three specific issues: (i) Problems in conceptual planning of multi-proposition utterances of the type proposed by Arbib (2005), Mithen (2005); (ii) The question of whether holophrastic protolanguage could have been processed by a special holistic mode, the precursor to a projected idiom mode in modern language; (iii) The implications for learning a holophrastic proto-lexicon in light of lexical constraints on word learning. Modern speakers only plan utterances in clause-sized units, and it is improbable that protolanguage speakers had more complex abilities. Moreover, the production and comprehension of idioms sheds no light on a putative holistic mode of language processing, since idioms are not processed in this way. Finally, innate constraints on learning lexical items preclude the types of word meanings proposed by proponents of holophrastic protolanguage.

One important difference between existing accounts of protolanguage lies in their assumptions on the semantic complexity of protolinguistic utterances. I bring evidence about the nature of linguistic communication to bear on the plausibility of these assumptions, and show that communication is fundamentally inferential and characterised by semantic uncertainty. This not only allows individuals to maintain variation in linguistic representation, but also imposes a selection pressure that meanings be reconstructible from context. I argue that protolanguage utterances had varying degrees of semantic complexity, and developed into complex language gradually, through the same processes of re-analysis and analogy which still underpin continual change in modern languages.

Early humans formed language units consisting of global and discrete dimensions of semiosis in dynamic opposition, or growth points. At some point, gestures gained the power to orchestrate actions, manual and vocal, with significances other than those of the actions themselves, giving rise to cognition framed in dual terms. However, our proposal emphasizes natural selection of joint gesture-speech, not gesture-first in language origin.

It is possible for a language to emerge with no direct linguistic history or outside linguistic influence. Al-Sayyid Bedouin Sign Language (ABSL) arose about 70 years ago in a small, insular community with a high incidence of profound prelingual neurosensory deafness. In ABSL, we have been able to identify the beginnings of phonology, morphology, syntax, and prosody. The linguistic elements we find in ABSL are not exclusively holistic, nor are they all compositional, but a combination of both. We do not, however, find in ABSL certain features that have been posited as essential even for a proto-language. ABSL has a highly regular syntax as well as word-internal compounding, also highly regular but quite distinct from syntax in its patterns. ABSL, however, has no discernable word-internal structure of the kind observed in more mature sign languages: no spatially organized morphology and no evident duality of phonological patterning.

Much of the debate concerning the question ''Was Protolanguage Holophrastic?'' assumes that protolanguage existed as a single, stable transitional form between communication systems akin to those of modern primates and human languages as we know them today. The present paper argues for a spectrum of protolanguages preceding modern languages emphasizing that (i) protospeech was intertwined with protosign and gesture; (ii) grammar emerged from a growing population of constructions; and (iii) an increasing protolexicon drove the emergence of phonological structure. This framework weakens arguments for the view that the earliest protolanguages were not holophrastic while advancing the claim that protolanguages became increasingly compositional over time en route to the emergence of true languages.

In dealing with the nature of protolanguage, an important formative factor in its development, and one that would surely have influenced that nature, has too often been neglected: the precise circumstances under which protolanguage arose. Three factors are involved in this neglect: a failure to appreciate radical differences between the functions of language and animal communication, a failure to relate developments to the overall course of human evolution, and the supposition that protolanguage represents a package, rather than a series of separate developments that sequentially impacted the communication of pre-humans. An approach that takes these factors into account is very briefly suggested.

This paper illustrates how external (or {\em social}) symbol grounding can be studied in simulations with large populations. We discuss how we can simulate language evolution in a relatively complex environment which has been developed in the context of the New Ties project. This project has the objective of evolving a cultural society and, in doing so, the agents have to evolve a communication system that is grounded in their interactions with their virtual environment and with other individuals. A preliminary experiment is presented in which we investigate the effect of a number of learning mechanisms. The results show that the social symbol grounding problem is a particularly hard one; however, we provide an ideal platform to study this problem.

Abstract: This paper compares the nascent theory of the'semiotic symbol'in cognitive science with its computational relative. It finds that the semiotic symbol as it is understood in recent practical and theoretical work does not have the resources to explain the role of symbols ...

It has long been asserted that the evolutionary path to spoken language was paved by manual-gestural behaviors, a claim that has been revitalized in response to recent research on mirror neurons. Renewed interest in the relationship between manual and vocal behavior draws attention to its development. Here, the pointing and vocalization of 16.5-month-old infants are reported as a function of the context in which they occurred. When infants operated in a referential mode, the frequency of simultaneous vocalization and pointing exceeded the frequency of vocalization-only and pointing-only responses by a wide margin. In a non-communicative context, combinatorial effects persisted, but in weaker form. Manual-vocal signals thus appear to express the operation of an integrated system, arguably adaptive in the young from evolutionary times to the present. It was speculated, based on reported evidence, that manual behavior increases the frequency and complexity of vocal behaviors in modern infants. There may be merit in the claim that manual behavior facilitated the evolution of language because it helped make available, early in development, behaviors that under selection pressures in later ontogenetic stages elaborated into speech.

A trait is of adaptive value if it confers a fitness advantage to its possessor. Thus adaptivness is an ahistorical identification of a trait affording some selective advantage to an agent within some particular environment. In results reported here we identify a trait within natural language discourse as having adaptive value by computing a trait/fitness covariance; the possession of the trait correlates with the replication success of the trait's possessor. We show that the trait covaries with fitness across multiple unrelated discursive groups. In our analysis the trait in question is a particular statistically derived word?n? context, that is, a word set. Variation of the word?sage is measured as the relative presence of the word set within a particular text, that is, the percentage of the text devoted to this set of words. Fitness is measured as the rate in which the text is responded to, or replicates, within an online environment. Thus we are studying the micro?volutionary dynamics of natural language discourse.

This paper explores a structure ubiquitous in grammar, the embedded tree, and develops a proposal for how such embedded structures played a fundamental role in the evolution of consonants and vowels. Assuming that linguistic capabilities emerged as a cognitive system from a simply reactive system and that such a transition required the construction of an internal mapping of the system body (cf. Cruse 2003), we propose that this mapping was determined through articulation and acoustics. By creating distinctions between articulators in the vocal tract or by acoustic features of sounds, and then embedding these distinctions, the various possible properties of consonants and vowels emerged. These embedded distinctions represent paradigmatic options for the production of sounds, which provide the basic building blocks for prosodic structure. By anchoring these embedded structures in the anatomy and physiology of the vocal tract, the evolution of phonology itself can be explained by extra-linguistic factors.

Most current modeling for evolution of communication still underplays or ignores the role of local action in spatialized environments: the fact that it is immediate neighbors with which one tends to communicate, and from whom one learns strategies or conventions of communication. Only now are the lessons of spatialization being learned in a related field: game-theoretic models for cooperation. In work on altruism, on the other hand, the role of spatial organization has long been recognized under the term `viscosity'.
Here we offer some simple simulations that dramatize the importance of spatialization for studies of both cooperation and communication, in each case contrasting (a) a model dynamics in which strategy change proceeds globally, and (b) a spatialized model dynamics in which interaction and strategy change both operate purely locally. Local action in a spatialized model clearly favors the emergence of cooperation. In the case of communication, spatialized models allow communication to arise and flourish where the global dynamics more typical in the literature make it impossible.
Simulations make a dramatic case for spatialized modeling, but analysis proves difficult. In a final section we outline some of the surprises of spatial dynamics but also some of the complexity facing attempts at deeper analysis.

To understand evolutionary aspects of communication is to understand the evolutionary development of the meaning relations between language and the world. Such meaning relations are established by the application of the interactive systems of semantic games. Subsumed under the evolutionary framework of repeated games, semantics in such games refers to the cases in which stable meanings survive populations of strategically interacting players. The viability of compositionality, common ground and salience in such evolutionary games is assessed. Foundationally, the discussion is rooted in Charles S. Peirce's pragmatist philosophy.

Previous simulation work on the evolution of communication has not shown how a large signal repertoire could emerge in situated agents. We present an artificial life simulation of agents, situated in a two-dimensional world, that must search for other agents with whom they can trade resources. With strong restrictions on which resources can be traded for others, initially non-communicating agents evolve/learn a signal system that describes the resource they seek and the resource they are willing to offer in return. A large signal repertoire emerges mainly through an evolutionary process. Agents whose production and comprehension abilities rely on a single mechanism fare best, although learning enables agents with separate mechanisms to achieve some measure of success. These results demonstrate that substantial signaling repertoires can evolve in situated multi-agent systems, and suggest that simulated social interactions such as trading may provide a useful context for further computational studies of the evolution of communication.

Different theories have been proposed for explaining the evolution of language. One of this maintains that gestural communication has been the precursor of human speech. Here we present a series of neurophysiological evidences that support this hypothesis. Communication by gestures, defined as the capacity to emit and recognize meaningful actions, may have originated in the monkey motor cortex from a neural system whose basic function was action understanding. This system is made by neurons of monkey’s area F5, named mirror neurons, activated by both execution and observation of goal-related actions. Recently, two new categories of mirror neurons have been described. Neurons of one category respond to the sound of an action, neurons of the other category respond to the observation of mouth ingestive and communicative actions. The properties of these neurons indicate that monkey's area F5 possesses the basic neural mechanisms for associating gestures and meaningful sounds as a pre-adaptation for the later emergence of articulated speech. The homology and the functional similarities between monkey area F5 and Broca’s area support this evolutionary scenario.