Language Evolution and Computation Bibliography

Recently in the Vocalize-to-Localize framework (a functional stance just started in the Interaction Studies 2004-2005 issues we edited, Abry et al., 2004), we addressed the unification of two grounding attempts concerning the syllable and the foot in language ontogeny. Can the movement time of the pointing strokes of a child be predicted from her babbling rhythm? The answer for 6 babies (6-18 months) was a 2:1 pointing-to-syllable ratio. Implications for the grounding of the first words within this Pointing Frame will be examined. More tentatively we will suggest that babbling for protophonology together with pointing for protosyntax pave the way to language.

The present article raises the so far unaddressed question of the neural foundations of the linguistic processes underlying grammaticalization operations. Two fundamental adaptive neurolinguistic mechanisms are presented: The cerebellar-induced Kalman gain reduction in linguistic processing, and the basal ganglionic re-regulation of cortical unification operations.

Abstract: According to Pragmatics, the speaker's intention in declarative speech is to complete or correct the hearer's belief. But the age at which children begin to produce this type of communication is prior to their success in the 'false belief'test. Here, after proposing ...

Recursion is considered to be one of the hallmarks of human communication. Many theories have been proposed on how this important feature might once have originated. This paper critically examines previously proposed models and posits a new and clearly defined hypothesis: recursion might originate from language users who try to reuse as much of their previously gained linguistic knowledge as possible. We support this claim by providing results of a multi-agent computer simulation in which the agents invent their own communication system encompassing a recursive syntactic category system.

We present evidence that, for English, ambiguity is an active factor in the choice of relativization strategy and that, in speech, prosody plays a role in resolution of ambiguity over the internal role of the relativized constituent. The evidence is based on (semi-)automatic analysis and comparison of automatically-parsed written and spoken portions of the British National Corpus (BNC, Leech, 1992) and of the prosodically-transcribed Spoken English Corpus (SEC, Taylor and Knowles, 1988). The results are evaluated with respect to a model of parsing complexity and syntactic disambiguation (Briscoe 1987, 2000) building on Combinatory Categorial Grammar (Steedman, 2000) and this model is in turn motivated by an evolutionary account of linguistic coevolutionary adaptation of the syntactic and phonological prosodic systems to a solution which minimizes processing cost. To our knowledge this is the first work which investigates linguistic adaptations aimed at reducing ambiguity while making testable predictions about linguistic organization.

In the general context of dynamics of social consensus, we study an agent based model for the competition between two socially equivalent languages, addressing the role of bilingualism and social structure. In a regular network, we study the formation of linguistic domains and their interaction across the boundaries. We also analyse the dynamics on a small world network and on a network with community structure. In all cases, a final scenario of dominance of one language and extinction of the other is obtained (dominance-extinction state). In comparison with the regular network, smaller times for extinction are found in the small world network. In the network with communities instead, the average time for extinction does not give a characteristic time for the dynamics, and metastable states are observed at all time scales.

5 to 6 million years ago, 3.5 megabase of DNA duplicated from the long arm of the X to create a Hominid-specific stratum on the Y short arm. Such events constitute saltations and this particular change is a candidate for the speciation event for Australopithecus. Within the transposed block a gene pair - Protocadherin X. and Protocadherin Y - has been subject to accelerated evolution, with 16 amino acid changes in the Y protein and five in the X. The latter are particularly significant in that they include the introduction of two sulphur containing cysteines, that are likely to have changed the function of the molecule, and are expressed in both males and females. The sequence changes are seen as secondary to chromosomal rearrangements on the Y. (four deletions and a paracentric inversion), the latter representing the initiating events in successive speciations, and the former representing the sexually selected phase of accommodation that establishes a new mate recognition system. The paracentric inversion, which has not been dated, is a candidate for the sapiens speciation event. Cerebral asymmetry (the torque), may have been introduced at a late stage, perhaps as recently as 160KYA.

It is argued that compositionality, hierarchy and recursion, generally acknowledged to be universal features of human languages, can be explained as being emergent properties of the complex dynamics governing the establishment and evolution of a language in a population of language users, mainly on an intra-generational time scale, rather than being the result of a genetic selection process leading to a specialized language faculty that imposes those features upon language or than being mainly a cross-generational cultural phenomenon. This claim is supported with results from a computational language game experiment in which a number of autonomous software agents bootstrap a common compositional and recursive language.

The causal correlations between human genetic variants and linguistic (typological) features could represent the mechanism required for gradual, accretionary models of language evolution. The causal link is mediated by the process of cultural transmission of language across generations in a population of genetically biased individuals. The particular case of Tone, ASPM and Microcephalin is discussed as an illustration. It is proposed that this type of genetically-influenced linguistic bias, coupled with a fundamental role for genetic and linguistic diversities, provides a better explanation for the evolution of language and linguistic universals.

The fact that human beings universally put much energy and conviction in reporting events in daily conversations demands an explanation. After having observed that the selection of reportable events is based on unexpectedness and emotion, we make a few suggestions to show how the existence of narrative behaviour can be consistent with the socio-political theory of the origin of language.

It is widely assumed that long-distance dependencies between elements are a unique feature of human language. Here we review recent evidence of long-distance correlations in sequences produced by non-human species and discuss two evolutionary scenarios for the evolution of human language in the light of these findings. Though applying their methodological framework, we conclude that some of Hauser, Chomsky and Fitch s central claims on language evolution are put into question to a different degree within each of those scenarios.

This paper presents a simulation study exploring the role of cultural transmission in intention sharing (the ability to establish shared intentions in communications). This ability has been argued to be human-unique, and a low level of it has deprived animals of the possibility of developing human language. Our simulation results show that the adequate level of this ability to trigger a communal language is not very high, and that cultural transmission can indirectly optimize the average level of this ability in the population. This work extends the current discussion on the human-uniqueness of some language-related abilities, and provides better understanding on the role of cultural transmission in language evolution.

This paper presents a simulation study to explore the role of naming game in social structure, which is nearly neglected by contemporary studies from statistical physics that mainly discuss the dynamics of naming game in predefined mean-field or complex networks. Our foci include the dynamics of naming game under some simpler, distance restrictions, and the origin and evolution of primitive social clusters as well as their languages under these restrictions. This study extends the current work on the role of social structure in language games, and provides better understanding on the self-organizing process of lexical conventionalization during cultural transmission.

We argue that studying grammaticalisation is useful to evolutionary linguists, if we abstract away from linguistic description to the underlying cognitive mechanisms. We set out a unified approach to grammaticalisation that allows us to identify these mechanisms, and argue that they could indeed be sufficient for the initial emergence of linguistic signal-meaning associations.

In holistic theories of protolanguage, a vital step is the fractionation process where holistic utter- ances are broken down into segments, and segments associated with semantic components. One problem for this process may be the occurrence of counterexamples to any segment-meaning connection. The actual abundance of such counterexamples is a contentious issue (Smith, 2006; Tallerman, 2007). Here I present calculations of the prevalence of counterexamples in model languages. It is found that counterexamples are indeed abundant, much more numerous than positive examples for any plausible holistic language.

One question above all divides evolutionary psychologists. Do humans have digital minds in an analog world Or should this idea be completely reversed? Is it that we have analog minds, although unlike our primate relatives we inhabit a digital world This paper is a contribution intended in the philosophical spirit of Botha s (2003) Unravelling the Evolution of Language.

It is widely assumed that, over their evolutionary history, languages increased in complexity from simple signals to protolanguages to complex syntactic structures. This papers investigates processes for increasing linguistic complexity while maintaining communicability across a pop- ulation. We assume that linguistic communicability is important for reliably exchanging infor- mation critical for coordination-based tasks. Interaction, needed for learning others languages and converging to communicability, bears a cost. There is a threshold of interaction (learning) effort beyond which convergence either doesn t pay or is practically impossible. Our central findings, established mainly through simulation, are: 1) There is an effort-dependent frontier of tractability for agreement on a language that balances linguistic complexity against linguis- tic diversity in a population. For a given maximum convergence effort either a) languages must be simpler, or b) their initial average communicability must be greater. Thus, if either conver- gence cost or high average communicability over time are important, then even agents who have the capability for using complex languages must not invent them from the start; they must start simple and grow. 2) A staged approach to increasing complexity, in which agents initially con- verge on simple languages and then use these to scaffold greater complexity, can outperform initially-complex languages in terms of overall effort to convergence. This performance gain improves with more complex final languages.

Over the last decade, computational models and simulations have been used to explore whether words could have initially become grounded and established in the earliest stages of language evolution through a process of self-organisation in a population. In this paper, a new model of this family is produced, with two major differences from previous models: the agents world consists of an infinite number of objects, and the agents categories have a flexible prototype structure. These changes result in a more realistic model, but also one which is more likely to fail. Simulation results revealed that coherent lexicons still emerged, but they were sensitive to certain model conditions, including the structure of the world.

Due to the nature of the subject, the field of language evolution has to rely largely on theoretical considerations. A coherent fundamental framework for approaching language evolution has to relate principles of evolution of complex traits with those governing the organization of cognitive processes, communication and natural language architecture. We suggest that by treating the language faculty as a complex trait with predefined functional interfaces, it is possible to delineate the evolutionary forces that have led to the emergence of natural language. We analyze embedding and recursion in communication, and propose a conceptual prerequisite for natural language and fully symbolic reference: a hierarchical way of conceptualization termed 'conceptual embedding' (the ability to nest concepts within concepts). We argue that parallel (multidimensional) interpretation and evolving abstract categories are the effects of conceptual embedding. We review and analyze relevant experiments in ethological literature and conclude that their results do not imply conceptual embedding in non-humans, which leads us to hypothesize that conceptual embedding may be a uniquely human trait. We go on to hypothesize that, initially, the selective force driving the development of the language faculty was towards enhanced conceptualization of reality that is functionally relevant in the absence of linguistic communication. According to this scenario, the invention of linguistic communication was a secondary event, dependent on conceptual embedding which supports the sophisticated conceptual underpinnings of linguistic meaning.

This paper presents a simulation model of self-organizing lexical networks. Its starting point is the notion of an association game in which the impact of varying community models is studied on the emergence of lexical networks. The paper reports on experiments whose results are in accordance with findings in the framework of the naming game. This is done by means of a multilevel network model in which the correlation of social and of linguistic networks is studied.

We demonstrate that polysemous links have a profound impact on the organization of the semantic graph, conforming it as a small-world network, based on the data from WordNet and A Thesaurus of Old English. We then show that the words with higher frequency and therefore with higher number of meanings construct the higher level of the hypernymy tree and this architecture is robust through the times. We suggest that our quantitative analysis of structural connection patterns in semantic networks provides insights into the functioning of neural architectures in the human brain. Furthermore, we discuss that vast quantities of variable and random utterances are converted into a small-world of fixed forms with fixed meanings by the cooperation between humans in the emergence of lexicon.

We present a series of studies investigating the formation, generative power, and evolution of toponyms (i.e. topographic names). The domain chosen for this project is the spatial concepts related to movement through the environment, one of the key sets of concepts to be grounded in autonomous agents. Concepts for spatial locations cannot be directly perceived and require representations built from interactions and inferred from ambiguous sensory data. A generative toponymic language game has been developed to allow the agents to interact, forming concepts for locations and spatial relations. The studies have shown that a grounded generative toponymic language may form and evolve in a population of agents interacting through language games. Initially, terms are grounded in simple spatial concepts directly experienced by the robots. The generative process then enables the robots to learn about and refer to locations beyond their direct experience, enabling concepts and toponyms to co-evolve.

The defining problem of animal signalling theory is how reliable communication systems remain stable. The problem comes into sharp focus when signals take an arbitrary form, as human words do. Many researchers, particularly those in evolutionary linguistics, assume that the handicap principle is the only recognised solution to this paradox, and hence conclude that the process that underpins reliability in humans must be exceptional. However, there are other examples of cheap yet reliable signals in nature, and corresponding processes that may explain such examples. This paper reviews these alternatives and concludes that by far the most likely explanation of the stability of human communication is our ability to assess individual reputation: we hold the threat of social exclusion against those who signal unreliably and hence keep their utterances reliable.

There are two possible sources of structure in language: biological evolution of the language faculty, or cultural evolution of language itself. Two recent models (Griffiths & Kalish, 2005; Kirby, Dowman, & Griffiths, 2007) make alternative claims about the relationship between innate bias and linguistic structure: either linguistic structure is largely determined by cultural factors (Kirby et al., 2007), with strength of innate bias being relatively unimportant, or the nature and strength of innate machinery is key (Griffiths & Kalish, 2005). These two competing possibilities rest on different assumptions about the learning process. We extend these models here to include a treatment of biological evolution, and show that natural selection for communication favours those conditions where the structure of language is primarily determined by cultural transmission.

General physical laws are evident as universal syntactic principles governing a computational system of the human language. The optimal space filling condition has to be satisfied in every system of efficient growth. This principle can be attested in syntax, exemplified as the Fibonacci (Fib)-patterns where each new term is the sum of the two that precede it. It will be shown that this rule accounts for the essential features of syntactic trees: limitations imposed on the number of arguments, and phase formation in derivations. The article provides a functional explanation of binary branching, labeling, and the properties of External and Internal Merge.

We show that artificial language evolution involves the interplay of two opposing forces: pres- sure toward simple representations imposed by the dynamics of collective learning, and pres- sure towards complex representations imposed by requirements of agents tasks. The push-pull of these two forces results in the emergence of a language that is balanced: simple but not too simple. We introduce the classification game to study the emergence of these balanced languages and their properties. Our agents use artificial neural networks to learn how to solve tasks, and a simple counting algorithm to simultaneously learn a language as a form-meaning mapping. We show that task-language coupling drives the simplicity-complexity balance, and that both compositional and holistic languages can emerge.

Language is typically argued (or assumed) to be an adaptive trait in the Homo lineage, and various specific selection pressures are offered to explain why language would have increased fitness in a population. However, it is incoherent to discuss language as a monolithic entity: the set of properties that comprise the full, complex language faculty almost certainly evolved independently, and any pressure that buys one of these properties does not necessarily entail the others. Some recent work on kin selection starts by discussing the evolution of speech, but then moves on to the selective value of the exchange of information without indicating how our ancestors got from vocalization to propositions. This is too large a leap, and more specific mechanisms must be proposed if the hypotheses are to be seriously considered.

We combine information theory and cross-situational learning to develop a novel metric for quantifying the degree of regularity in the mappings between signals and meanings that can be inferred from exposure to language in context. We illustrate this metric using the results of two artificial language learning experiments, which show that learners are sensitive, with a high level of individual variation, to systematic regularities in the input. Analysing language using this measure of regularity allows us to explore in detail how language learning and language use can both generate linguistic variation, leading to language change, and potentially complexify language structure, leading to qualitative language evolution.

The ability to interpret, conceive and learn composite meaning is a pre- requisite for language use. Any computational model of the emergence and evolution of grammar needs to consider the emergence and evolution of such meaning. It needs to explain how interpretation deals with semantic ambi- guity, how rich meaning is conceptualised, and how it is acquired and con- ventionalised. Various computational models have been proposed that deal with one or a few of these aspects. It is however hard to integrate them given the diverse and often hard to align underlying conceptual and computational metaphors and paradigms. To remedy this we propose a constraint-based model of compositional semantics which affords a uniform and grounded treatment of its interpretation, its conceptualisation and its acquisition.

This paper shows how experiments on artificial language evolution can provide highly relevant results for important debates in linguistic theories. It reports on a series of experiments that in- vestigate how semantic roles can emerge in a population of artificial embodied agents and how these agents can build a network of constructions. The experiment also includes a fully oper- ational implementation of how event-specific participant-roles can be fused with the semantic roles of argument-structure constructions and thus contributes to the linguistic debate on how the syntax-semantics interface is organized.

In current literature, a number of standard lines of evidence reemerge in support of the hypothesis that the initial, bootstrapping stage of the evolution of language was gestural. However, one specific feature of gestural communication consistent with this hypothesis has been given surprisingly little attention. The visual modality makes gestural signals more secret than vocal signals (lack of broadcast transmission). The high relevance of secrecy is derived from the fundamental constraint on language evolution: the transfer of honest messages itself is a form of cooperation, and therefore not a naturally evolutionarily stable strategy. Consequently, greater secrecy of gestural communication constitutes a potentially important factor that should not fail to be represented in more comprehensive models of the emergence of protolanguage.

We examine the social prerequisites for symbolic communication by studying a language game embedded within a signaling game, in which cooperation is possible but unenforced. Despite incentives to cheat, and even with persistent cheating, the lateral inhibition dynamics commonly used in language game models remain resilient, as long as sufficient mechanisms are in place to detect deceit. However, unfairly antagonistic strategies can undermine lexical convergence. Symbolic communication, and hence human language, requires a delicate balance between restrained deception and revocable trust, but unconditional cooperation is unnecessary.

Scaling up the complexity of a language game brings about a towering scale-up in the uncer- tainty the agents are faced with when acquiring (lexical) form-meaning associations. The two most prominent assumptions influencing the uncertainty in models on word meaning concern (1) meaning transfer and (2) whether a form can be associated with only one part of meaning or any subset of parts of meaning. If meaning has internal structure (e.g. sets of attributes) this second assumption amounts to whether a form can be associated with only one attribute, giving rise to linear uncertainty, or any subset of attributes, resulting in exponential uncertainty. We first present a short overview of different models that each tried to tackle at least one of these assumptions. We propose a new model borrowing ideas from many of these models that can handle the exponential increase in uncertainty when removing both assumptions and allows scaling towards very large meaning spaces (i.e. worlds).

The iterated learning model (ILM), in which a language comes about via communication pressures exerted over successive generations of agents, has attracted much attention in recent years. Its importance lies in the focus on cultural emergence as opposed to biological evolution. The ILM simplifies a compositional language as the compression of an object space, motivated by a poverty of stimulus ''as not all objects in the space will be encountered by an individual in its lifetime. However, in the original ILM, every agent magically has a complete understanding of the surrounding object space, which weakens the relevance to natural language evolution. In this paper, we define each agent s meaning space as an internal self-organising map, allowing it to remain personal and potentially unique. This strengthens the parallels to real language as the agent s omniscience and mind-reading abilities that feature in the original ILM are removed. Additionally, this improvement motivates the compression of the language through a poverty of memory as well as a poverty of stimulus. Analysis of our new implementation shows maintenance of a compositional (structured) language. The effect of a (previously-implicit) generalisation parameter is also analysed; when each agent is able to generalise over a larger number of objects, a more stable compositional language emerges.