Language Evolution and Computation Bibliography

This paper investigates the formation of color categories and color naming in a population of agents. The agents perceive and categorize color stimuli, and try to communicate about these perceived stim- uli. While doing so they adapt their internal representations to be more successful at conveying color meaning in future interactions. The agents have no access to global information or to the representa- tions of other agents; they only exchange word forms. The factors driving the population coherence are the shared environment and the interactions. The experiments show how agents can form a coherent lexicon of color terms and -particularly- how a coherent color categorization emerges through these linguistic interactions. The results are interpreted in the light of theories describing and explaining universal tendencies in human color categorization and color naming. At the same time, the experiments confirm aspects of the theories of Luc Steels who views language as a complex dynamic system, arising from selforganization and cultural interactions.

Recent work in the field of computational evolutionary linguistics suggests that the dynamics arising from the cultural evolution of language can explain the emergence of syntactic structure. We build on this work by introducing a model of language acquisition based on the Minimum Description Length Principle. Our experiments show that compositional syntax is most likely to occur under two conditions specific to hominids: (i) A complex meaning space structure, and (ii) the poverty of the stimulus.

Artificial Life models have consistently implemented communication as an exchange of signals over dedicated and functionally isolated channels. I argue that such a feature prevents models from providing a satisfactory account of the origins of communication and present a model in which there are no dedicated channels. Agents controlled by neural networks and equipped with proximity sensors and wheels are presented with a co-ordinated movement task. It is observed that functional, but non-communicative, behaviours which evolve in the early stages of the simulation both make possible, and form the basis of, the communicative behaviour which subsequently evolves.

This paper investigates the development of experience-based meaning creation and explores the problem of establishing successful communication systems in a population of agents. The aim of the work is to investigate how such systems can develop, without reliance on phe- nomena not found in actual human language learning, such as the explicit transmission of meaning or the provision of reliable error feedback to guide learning. Agents develop individual, distinct meaning structures, and although they can communicate despite this, communicative success is closely related to the proportion of shared lexicalised meaning, and the communicative systems have a large degree of redundant synonymy.

Oliphant [5,6] contends that language is the only naturally-occurring, learned symbolic communication system, because only humans can accurately observe meaning during the cultural transmission of communication. This paper outlines several objections to Oliphant's argument. In particular, it is argued that the learning biases necessary to support learned symbolic communication may not be common and that the speed of cultural convergence during cultural evolution of communication may be a key factor in the evolution of such learning biases.

Turkel [16] studies a computational model in which agents try to establish communication. It is observed that over the course of evolution, initial plasticity is significantly nativised. This result supports the idea that innate language knowledge is explained by the Baldwin effect [2][14]. A more biologically plausible computational model, however, reveals the result is unsatisfactory. Implications of this new representation system in language evolution are discussed with a consideration of the Baldwin effect.

In this paper we explore the similarities between a mathematical model of language evolution and several A-life simulations. We argue that the mathematical model makes some problematic simplifications, but that a combination with computational models can help to adapt and extend existing language evolution scenario's.

This paper presents an Artificial Life and Neural Network (ALNN) model for the evolution of syntax. The simulation methodology provides a unifying approach for the study of the evolution of language and its interaction with other behavioral and neural factors. The model uses an object manipulation task to simulate the evolution of language based on a simple verb-noun rule. The analyses of results focus on the interaction between language and other non-linguistic abilities, and on the neural control of linguistic abilities. The model shows that the beneficial effects of language on non-linguistic behavior are explained by the emergence of distinct internal representation patterns for the processing of verbs and nouns.

Many models, computational or not, exist that describe the acquisition of speech: they all rely on the pre-existence of some sort of linguistic structure in the input, i.e. speech itself. Very few address the question of how this coherence and structure appeared. We try here to give a solution concerning syllable systems. We propose an operational model that shows how a society of robotic of agents, endowed with a set of non-linguistically specific motor, perceptual, cognitive and social constraints (some of them are obstacles whereas others are opportunities), can collectively build a coherent and structured syllable system from scratch. As opposed to many existing abstract models of the origins of language, as few shortcuts as possible were taken in the way the constraints are implemented. The structural properties of the produced sound systems are extensively studied under the light of phonetics and phonology and more broadly language theory. The model brings more plausibility in favor of theories of language that defend the idea that there needs no innate linguistic specific abilities to explain observed regularities in world languages.

We analyze a general model of multi-agent communication in which all agents learn to communicate simultaneously to a message board. We show that the communicating multi-agent system is equivalent to a Mealy finite state machine whose states are determined by the agents' usage of the learned language. Increasing the language size increases the number of possible states in the Mealy machine, and can improve the performance of the multi-agent system. We introduce the term \em semantic density to describe the average number of meanings assigned to each word of a language. Using semantic density, a simple rule is presented that provides a pessimistic estimate of the minimum language size that should be used for any multi-agent problem in which the agents communicate simultaneously. Simulations on a version of the predator-prey pursuit problem, a simplified version of problems seen in warfare scenarios, validate these predictions. The communicating predators evolved using a genetic algorithm perform significantly better than all previous work on similar preys.

A unified connectionist model of the perceptual magnet effect (the perceptual warping of vowels) is proposed, and relies on the concept of population coding in neural maps. Unlike what has been often stated, we claim that the imprecision of the classical sum of vectors coding/decoding scheme is not a drawback and can account for psychological observations. Furthermore, we show that coupling these neural maps allows the formation of vowel systems, which are shared symbolic systems, from initially continuous and uniform perception and production. This has important consequences for existing theories of phonetics.

This paper presents a model of the origins of syllable systems that brings plausibility to the theory which claims that language learning, and in particular phonological acquisition, needs not innate linguistically specific information, as believed by many researchers of the Chomskyan school, but is rather made possible by the interaction between general motor, perceptual, cognitive and social constraints through a self-organizing process. The strategy is to replace the question of acquisition in a larger and evolutionary (cultural) framework: the model addresses the question of the origins of syllable systems (syllables are the major phonological units in speech). It is based on the artificial life methodology of building a society of agents, endowed with motor, perceptual and cognitive apparati that are generic and realistic. We show that agents effectively build sound systems and how these sound systems relate to existing human sound systems. Results concerning the learnability of the produced sound systems by fresh/baby agents are detailed: the critical period effect and the artificial language effect can effectively be predicted by our model. The ability of children to learn sound systems is explained by the evolutionary history of these sound systems, which were precisely shaped so as to fit the ecological niche formed by the brains and bodies of these children, and not the other way around (as advocated by Chomskyan approaches to language).

A computational model of the origins of syllables systems is presented : a society of robotic agents endowed with realistic motor, perceptual and cognitive apparati is shown to build from scratch shared syllable systems in a decentralized manner. Furthermo re, these systems share many structural properties with those of human languages.

This paper presents a series of experiments in which two mobile robots develop a shared lexicon of which the meaning is grounded in the real world. The experiments investigate the impact of non-verbal communication on lexicon formation. Non-verbal communication is used to establish joint attention or to evaluate feedback. The experiments implement adaptive language games in which two agents try to communicate some real world object. When the agents fail, they can adapt their memory in order to improve performance on future occasions. As the experimental results show, the quality of the evolved lexicon is better when feedback is used rather than joint attention.

This research focuses on the artificial creation of the conditions that were necessary or facilitated language in its evolution. I propose a model for a system that has the capability to allow a complex communication protocol to evolve. In human language learning, the adult humans have already mastered the language and use their knowledge to teach infant humans. 50.000 B.C. (or whenever) when there were no adult masters of language, what conditions were necessary for the language-less homo sapiens to start developing the first language? Can machines evolve a language in a similar manner across generations? This research deals with facilitating the genesis of a communication system using evolutionary computation and reinforcement learning when initially none of the conspirators have mastered the system.

Individuals affected with developmental disorders of speech and language have substantial difficulty acquiring expressive and/or receptive language in the absence of any profound sensory or neurological impairment and despite adequate intelligence and opportunity. Although studies of twins consistently indicate that a significant genetic component is involved, most families segregating speech and language deficits show complex patterns of inheritance, and a gene that predisposes individuals to such disorders has not been identified. We have studied a unique three-generation pedigree, KE, in which a severe speech and language disorder is transmitted as an autosomal-dominant monogenic trait. Our previous work mapped the locus responsible, SPCH1, to a 5.6-cM interval of region 7q31 on chromosome 7 (ref. 5). We also identified an unrelated individual, CS, in whom speech and language impairment is associated with a chromosomal translocation involving the SPCH1 interval. Here we show that the gene FOXP2, which encodes a putative transcription factor containing a polyglutamine tract and a forkhead DNA-binding domain, is directly disrupted by the translocation breakpoint in CS. In addition, we identify a point mutation in affected members of the KE family that alters an invariant amino-acid residue in the forkhead domain. Our findings suggest that FOXP2 is involved in the developmental process that culminates in speech and language.

Human biological and cultural evolution are closely linked to technological innovations. Direct evidence for tool manufacture and use is absent before 2.5 million years ago (Ma), so reconstructions of australopithecine technology are based mainly on the behavior and anatomy of chimpanzees. Stone tool technology, robust australopithecines, and the genus Homo appeared almost simultaneously 2.5 Ma. Once this adaptive threshold was crossed, technological evolution was accompanied by increased brain size, population size, and geographical range. Aspects of behavior, economy, mental capacities, neurological functions, the origin of grammatical language, and social and symbolic systems have been inferred from the archaeological record of Paleolithic technology.

Universal grammar specifies the mechanism of language acquisition. It determines the range of grammatical hypothesis that children entertain during language learning and the procedure they use for evaluating input sentences. How universal grammar arose is a major challenge for evolutionary biology. We present a mathematical framework for the evolutionary dynamics of grammar learning. The central result is a coherence threshold, which specifies the condition for a universal grammar to induce coherent communication within a population. We study selection of grammars within the same universal grammar and competition between different universal grammars. We calculate the condition under which natural selection favors the emergence of rule-based, generative grammars that underlie complex language.

The costly signaling hypothesis proposes that animal signals are kept honest by appropriate signal costs. We show that to the contrary, signal cost is unnecessary for honest signaling even when interests conflict. We illustrate this principle by constructing examples of cost-free signaling equilibria for the two paradigmatic signaling games of Grafen (1990) and Godfray (1991). Our findings may explain why some animal signals use cost to ensure honesty whereas others do not and suggest that empirical tests of the signaling hypothesis should focus not on equilibrium cost but, rather, on the cost of deviation from equilibrium. We use these results to apply costly signaling theory to the low-cost signals that make up human language. Recent game theoretic models have shown that several key features of language could plausibly arise and be maintained by natural selection when individuals have coincident interests. In real societies, however, individuals do not have fully coincident interests. We show that coincident interests are not a prerequisite for linguistic communication, and find that many of the results derived previously can be expected also under more realistic models of society.

Sequential learning plays a role in a variety of common tasks, such as human language processing, animal communication, and the learning of action sequences. In this article, we investigate sequential learning in non-human primates from a comparative perspective, focusing on three areas: the learning of arbitrary, fixed sequences; statistical learning; and the learning of hierarchical structure. Although primates exhibit many similarities to humans in their performance on sequence learning tasks, there are also important differences. Crucially, non-human primates appear to be limited in their ability to learn and represent the hierarchical structure of sequences. We consider the evolutionary implications of these differences and suggest that limitations in sequential learning may help explain why non-human primates lack human-like language.

Language is an apparent miracle. Children master it with exceptional ease, while simultaneously struggling to walk, hold a fork, and recognize that others have thoughts and emotions that differ from their own. They perform, with near perfection, mental computation and generalizations about language which are virtually impossible for state of the art computers. They grasp the tree-like phrase structure of language even though their parents have never taught them, and most probably couldn't even if they wanted to (such properties of language are not the stuff of school education). And children babble on about the present, past, and future, creating imaginary worlds that no one but they can see.

Language is a biological trait that radically changed the performance of one species and the appearance of the planet. Understanding how human language came about is one of the most interesting tasks for evolutionary biology. Here we discuss how natural selection can guide the emergence of some basic features of human language, including arbitrary signs, words, syntactic communication and grammar. We show how natural selection can lead to the duality of patterning of human language: sequences of phonemes form words; sequences of words form sentences. Finally, we present a framework for the population dynamics of grammar acquisition, which allows us to study the cultural evolution of grammar and the biological evolution of universal grammar.

Grammar is the computational system of language. It is a set of rules that specifies how to construct sentences out of words. Grammar is the basis of the unlimited expressibility of human language. Children acquire the grammar of their native language without formal education simply by hearing a number of sample sentences. Children could not solve this learning task if they did not have some pre-formed expectations. In other words, children have to evaluate the sample sentences and choose one grammar out of a limited set of candidate grammars. The restricted search space and the mechanism which allows to evaluate the sample sentences is called universal grammar. Universal grammar cannot be learned; it must be in place when the learning process starts. In this paper, we design a mathematical theory that places the problem of language acquisition into an evolutionary context. We formulate equations for the population dynamics of communication and grammar learning. We ask how accurate children have to learn the grammar of their parents' language for a population of individuals to evolve and maintain a coherent grammatical system. It turns out that there is a maximum error tolerance for which a predominant grammar is stable. We calculate the maximum size of the search space that is compatible with coherent communication in a population. Thus, we specify the conditions for the evolution of universal grammar.

A distinctive feature of all human languages is the diverse and arbitrary nature of the sign (signifier). This can be interpreted as stating that the mapping between signals and referents is established by convention rather than by functional constraints. This property of the sign provides for a great deal of linguistic flexibility and is a key component of symbolic communication. Game theoretic models to describe signal imitation are investigated with a view to understanding how non-arbitrary (indexical) animal-style signals might 'evolve' culturally into diverse, arbitrary signs. I explore the evolutionary hypothesis that private, arbitrary signs emerge as a result of selective imitation within a socially structured population. Once arbitrary signs have emerged, they contribute towards greater assortative interactions among individuals using a shared sign system. In natural populations, the models for imitation will very often be close kin. Hence, kinship provides one mechanism for the creation of true symbols. An imitation-structured population can support many more sign systems than an equivalent non-structured population and is one in which symbols become the dominant force in assortative interactions.

In the research described here we extend past computational investigations of animal signaling by studying an artificial world in which a population of initially noncommunicating agents evolves to communicate about food sources and predators. Signaling in this world can be either beneficial (e.g., warning of nearby predators) or costly (e.g., attracting predators or competing agents). Our goals were twofold: to examine systematically environmental conditions under which grounded signaling does or does not evolve, and to determine how variations in assumptions made about the evolutionary process influence the outcome. Among other things, we found that agents warning of nearby predators were a common occurrence whenever predators had a significant impact on survival and signaling could interfere with predator success. The setting most likely to lead to food signaling was found to be difficult-to-locate food sources that each have relatively large amounts of food. Deviations from the selection methods typically used in traditional genetic algorithms were also found to have a substantial impact on whether communication evolved. For example, constraining parent selection and child placement to physically neighboring areas facilitated evolution of signaling in general, whereas basing parent selection upon survival alone rather than survival plus fitness measured as success in food acquisition was more conducive to the emergence of predator alarm signals. We examine the mechanisms underlying these and other results, relate them to existing experimental data about animal signaling, and discuss their implications for artificial life research involving evolution of communication.

Abstract: In order to understand the evolutionary pathway to the capability for language, we must first clearly understand the functional capabilities that the child brings to the task of language acquisition. Behavioral studies provide insight into infants' ability to extract ...

In this paper an experiment is presented in which two mobile robots develop a shared lexicon of which the meanings are grounded in the real world. The robots start without a lexicon nor shared meanings and play language games in which they generate new meanings and negotiate words for these meanings. The experiment tries to find the minimal conditions under which verbal communication may begin to evolve. The robots are autonomous in terms of computing and cognition, but they are otherwise far simpler than most, if not all animals. It is demonstrated that a lexicon nevertheless can be made to emerge even though there are strong limits on the size and stability of this lexicon.

Zipf's law states that the frequency of a word is a power function of its rank. The exponent of the power is usually accepted to be close to (-)1. Great deviations between the predicted and real number of different words of a text, disagreements between the predicted and real exponent of the probability density function and statistics on a big corpus, make evident that word frequency as a function of the rank follows two different exponents, \approx (-)1 for the first regime and \approx (-)2 for the second. The implications of the change in exponents for the metrics of texts and for the origins of complex lexicons are analyzed.

Language is the most important evolutionary invention of the last few million years. How human language evolved from animal communication is a challenging question for evolutionary biology. In this paper we use mathematical models to analyze the major transitions in language evolution. We begin by discussing the evolution of coordinated associations between signals and objects in a population. We then analyze word-formation and its relationship to Shannon's noisy coding theorem. Finally, we model the population dynamics of words and the adaptive emergence of syntax.

This paper describes different types of models for the evolution of communication and language. It uses the distinction between signals, symbols, and words for the analysis of evolutionary models of language. In particular, it show how evolutionary computation techniques, such as Artificial Life, can be used to study the emergence of syntax and symbols from simple communication signals. Initially, a computational model that evolves repertoires of isolated signals is presented. This study has simulated the emer- gence of signals for naming foods in a population of foragers. This type of model studies communication systems based on simple signal-object associations. Subsequently, models that study the emergence of grounded symbols are discussed in general, including a detailed description of a work on the evolution of simple syntactic rules. This model focuses on the emergence of symbol-symbol relationships in evolved languages. Finally, computational models of syntax acquisition and evolution are discussed. These different types of computational models provide an operational definition of the signal/symbol/word distinction. The simulation and analysis of these types of models will help understanding the role of symbols and symbol acquisition in the origin of language.

Describes an attempt to cast several abstract properties of natural languages in the framework of Kauffman's (1993, 1995) random Boolean nets (RBN). The properties are complexity, interconnectedness, stability, diversity, and underdeterminedness. A language is modeled as a Boolean net attractor. (Groups of) net nodes are linguistic principles or parameters as posited by Chomskyan theory, according to which the language learner sets parameters to appropriate values on the basis of very limited experience of the language. The setting of one parameter can have a complex effect on the settings of others. A RBN is generated to find an attractor. A state from this attractor is degraded, which represents the degenerate input of language to the learner, and this state is then input to a net with the same connectivity and activation functions as the original net to see whether it converges on the same attractor. Many nets degenerate into attractors representing complete uncertainty. Others settle at intermediate levels of uncertainty, and some manage to overcome the incompleteness of input and converge on attractors identical to that from which the original inputs were (de)generated. Finally, an attempt was made to select a population of such successful nets, using a genetic algorithm where fitness was correlated with an ability to acquire several different languages faithfully. This has so far proved impossible, supporting the Chomskyan suggestion that the human language acquisition capacity is not the outcome of natural selection.

A computationally implemented model of the transmission of linguistic behavior over time is presented. In this iterated learning model (ILM), there is no biological evolution, natural selection, nor any measurement of the success of the agents at communicating (except for results-gathering purposes). Nevertheless, counter to intuition, significant evolution of linguistic behavior is observed. From an initially unstructured communication system (a protolanguage), a fully compositional syntactic meaning-string mapping emerges. Furthermore, given a nonuniform frequency distribution over a meaning space and a production mechanism that prefers short strings, a realistic distribution of string lengths and patterns of stable irregularity emerges, suggesting that the ILM is a good model for the evolution of some of the fundamental features of human language.

This paper addresses Narrative Intelligence from a bottom up, Artificial Life perspective. First, different levels of narrative intelligence are discussed in the context of human and robotic story-tellers. Then, we introduce a computational framework which is based on minimal definitions of stories, story-telling and autobiographic agents. An experimental test-bed is described which is applied to the study of story-telling, using robotic agents as examples of situated, autonomous minimal agents. Experimental data are provided which support the working hypothesis that story-telling can be advantageous, i.e. increases the survival of an autonomous, autobiographic, minimal agent. We conclude this paper by discussing implications of this approach for story-telling in humans and artifacts.

This paper is an attempt to construct a programmatic framework for the evolution of human language. First, we pres- ent a novel characterization of language, which is based on some of the most recent research results in linguistics. As these results suggest, language is best characterized as a specialized communication system, dedicated to the expres- sion of a surprisingly constrained set of meanings. This characterization calls for an account of the evolution of lan- guage in terms of the interaction between cultural and genetic evolution. We develop such an evolutionary model on the basis of the mechanism of culturally-driven genetic assimilation. As we show, a careful analysis of the diverse effects of this mechanism derives some of the most crucial properties of the evolved linguistic capacity as a specific, functional communication system.

Human language may be described as a complex network of linked words. In such a treatment, each distinct word in language is a vertex of this web, and interacting words in sentences are connected by edges. The empirical distribution of the number of connections of words in this network is of a peculiar form that includes two pronounced power-law regions. Here we propose a theory of the evolution of language, which treats language as a self-organizing network of interacting words. In the framework of this concept., we completely describe the observed word web structure without any fitting. We show that the two regimes in the distribution naturally emerge from the evolutionary dynamics of the word web. It follows front our theory that the size of the core part of language, the 'kernel lexicon', does not vary as language evolves.

Words in human language interact in sentences in non-random ways, and allow humans to construct an astronomic variety of sentences from a limited number of discrete units. This construction process is extremely fast and robust. The co-occurrence of words in sentences reflects language organization in a subtle manner that can be described in terms of a graph of word interactions. Here, we show that such graphs display two important features recently found in a disparate number of complex systems. (i) The so called small-world effect. In particular, the average distance between two words, d (i.e. the average minimum number of links to be crossed from an arbitrary word to another), is shown to be d approximate to 2-3, even though the human brain can store many thousands. (ii) A scale-free distribution of degrees. The known pronounced effects of disconnecting the most connected vertices in such networks can be identified in some language disorders. These observations indicate some unexpected features of language organization that might reflect the evolutionary and social history of lexicons and the origins of their flexibility and combinatorial nature.

The language acquisition period in humans lasts about 13 years. After puberty it becomes increasingly difficult to learn a language. We explain this phenomenon by using an evolutionary framework. We present a dynamical system describing competition between language acquisition devices, which differ in the length of the learning period. There are two selective forces that play a role in determining the critical learning period: (i) having a longer learning period increases the accuracy of language acquisition; (ii) learning is associated with certain costs that affect fitness. As a result, there exists a limited learning period which is evolutionarily stable. This result is obtained analytically by means of a Nash equilibrium analysis of language acquisition devices. Interestingly, the evolutionarily stable learning period does not maximize the average fitness of the population.

The lexical matrix is an integral part of the human language system. It provides the link between word form and word meaning. A simple lexical matrix is also at the center of any animal communication system, where it defines the associations between form and meaning of animal signals. We study the evolution and population dynamics of the lexical matrix. We assume that children learn the lexical matrix of their parents. This learning process is subject to mistakes: (i) children may not acquire all lexical items of their parents (incomplete learning); and (ii) children might acquire associations between word forms and word meanings that differ from their parents' lexical items (incorrect learning). We derive an analytic framework that deals with incomplete learning. We calculate the maximum error rate that is compatible with a population maintaining a coherent lexical matrix of a given size. We calculate the equilibrium distribution of the number of lexical items known to individuals. Our analytic investigations are supplemented by numerical simulations that describe both incomplete and incorrect learning, and other extensions.

Natural selection is claimed to be the only way to explain complex design. The same assumption has also been held for language. However, sciences of complex-ity have shown, from a wide range of domains, the existence of a clear alternative: self-organisation, spontaneous patterns of order arising from chaos. According to this view, design derives from internal factors (dynamic interaction of the ele-ments within the system) rather than from adaptation to the environment by means of selection. This paper aims to apply sciences of complexity to language origins; it shows that preexisting and well established ideas can be rethought ac-cording to such a view. The main objective of the paper is to illustrate the new and promising horizons that complexity could open as regards the origins of the most specific property of human beings.

[From SWAP.] The most common word length in the lexicon lies in the middle of the total range. The shortest words -- light V or CV monosyllables -- are necessarily few because a cross-product of the consonantal and vocalic phonemes generates only a small number of combinations. As length increases, the number of possible forms explodes, but fewer and fewer are actually used in any given language. Experimental and computational studies relate the sparsity of long forms to the fact that the likelihood of forms as determined by a stochastic parse decreases with length. This effect occurs because long forms have more subparts than short ones, and the likelihood of any given subpart is always less than 1.0. (Coleman and Pierrehumbert 1997, Frisch et al in press). The disadvantage that long forms have in achieving a high well-formedness score is a distinct phenomenon from the tendency of individual long words to have low token frequencies (though there may of course be some deep relationship between these characteristcs). In English, a morphologically impoverished language, the most common type of word is the disyllable.

[Notes.] This is a Monte Carlo simulation on the learnability of phonological constraints, based on the idea that the phonological grammar needs to be shared even if individuals have different words in their vocabularies.

A computational model of emergent syllable systems is developed based on a set of functional constraints on syllable systems and the assumption that language structure emerges through cumulative change over time. The constraints were derived from general communicative factors as well as from the phonetic principles of perceptual distinctiveness and articulatory ease. Through evolutionary optimization, the model generated mock vocabularies optimized for the given constraints. Several simulations were run to understand how these constraints might define the emergence of universals and variation in complex sound systems. The predictions were that (1) CV syllables would be highly frequent in all vocabularies evolved under the constraints; (2) syllables with consonant clusters, consonant codas, and vowel onsets would occur much less frequently; (3) a relationship would exist between the number of syllable types in a vocabulary and the average word length in the vocabulary; (4) different syllable types would emerge according to, what we termed, an iterative principle of syllable structure and their frequency would be directly related to their complexity; and (5) categorical differences would emerge between vocabularies evolved under the same constraints. Simulation results confirmed these predictions and provided novel insights into why regularities and differences may occur across languages. Specifically, the model suggested that both language universals and variation are consistent with a set of functional constraints that are fixed relative to one another. Language universals reflect underlying constraints on the system and language variation represents the many different and equally-good solutions to the unique problem defined by these constraints.

Creole languages are often a point of contention for theoretical linguistics. Broadly defined, creoles result from an amalgamation of two or more languages, when speakers of differing mother tongues need rudimentary communication during economic or social transactions. Creolization occurs if the invented system becomes the native language of the speech community. There are several hypotheses for how biological linguistic properties and social contact each bear on the formation of creoles; however, until recently, no reliable method for testing these complex interactions existed. Implementing SWARM 2.1.1, the current model consists of a multiagent population drawn from historical records of Surinamese sugar cane plantations. Each agent in this artificial society is endowed with a demographic profile and linguistic parameters. Three experiments using the SWARM model are described. The results provide viable motivation for advancing a sociogenetic solution for the emergence of prototypical creole languages.

Integration and grounding are key AI challenges for human-robot dialogue. The author and his team are tackling these issues using language games and have experimented with them on progressively more complex platforms. The results of their work show that language games are a useful way to both understand and design human-robot interaction.

Human language has enabled our species to exchange information and to formulate ideas; understanding how human linguistic faculties evolved is one of the great challenges in evolutionary theory. Studies of the evolution of human language can be broadly separated into two types of approaches: those that consider the (e.g. phylogenetic) relationships between existing languages and their common ancestors; and those that try to understand the evolution of the human language capacity itself. For the latter case, Martin Nowak and co-workers have now shown that evolutionary game theory provides a framework in which the evolution of linguistic elements, such as word formation and syntax, can be investigated. These recent studies show that natural selection will favour the evolution of such `human' linguistic elements from simple animal communication if they enable more reliable exchange of relevant, that is fitness-enhancing, information.

This article recalls Cajal's brief mention of consciousness in the Textura as a function of the human brain quite distinct from reflex action, and discusses the view that human consciousness may share aspects of 'animal awareness' with other species, but has its unique form because humans possess language. Three ingredients of a theory of the evolution of human consciousness are offered: the view that a pres of intended activity is necessarily formed in the brain of a human that communicates in a human way; the notion that such a pres constitutes consciousness; and a new theory of the evolution of human language based on the mirror system of monkeys and the role of communication by means of hand gestures as a stepping-stone to speech.

Over the last decades, a considerable body of evidence has accumulated that speakers have detailed phonetic knowledge of a type which is not readily modelled using the categories and categorical rules of phonological theory. One line of evidence is systematic di-erences between languages in -ne details of pronunciation. For example, it is known that Spanish and English di-er systematically in the exact formant patterns typical of their point vowels (Bradlow 1995). Canadian French di-ers from both Canadian English and European French in the distribution of VOT times of voiced and voiceless stops (Caramazza and Yeni-Komshian, 1974). These are just two of many examples, with more reviewed in Pierrehumbert (in press) and Pierrehumbert et al. (in press); at this point, it is not possible to point to a single case in which analogous phonemes in two di-erent languages display exactly the same phonetic targets and the same pattern of phonetic variation in di-erent contexts. Exact phonetic targets and patterns of variation must accordingly be learned during the course of language acquisition. The usage-based framework readily accomodates such ndings by proposing that mental representations of phonological targets and patterns are gradually built up through experience with speech.

This book addresses universal tendencies of human vowel systems from the point of view of self-organization. It uses computer simulations to show that the same universal tendencies found in human languages can be reproduced in a population of artificial agents. These agents learn and use vowels with human-like perception and production, using a learning algorithm that is cognitively plausible. The implications of these results for the evolution of language are then explored.

This dissertation provides an Optimality-Theoretic (OT: Prince & Smolensky 1993, McCarthy & Prince 1993) account of the evolution of the English obstruent inventory over an extended period of time, starting with Proto-Indo-European and ending with Early New English. By modeling the mechanism through which the PIE [voice] distinction in stops emerged as a Germanic [spread glottis] distinction, the analysis lends support to a recently established claim that [spread glottis], rather than [voice], is distinctive in stops in the majority of the Germanic languages (Iverson & Salmons 1995, Jessen 1998) OT, a non-derivational framework which provides a model of Universal Grammar (UG) significantly different from the earlier rule-based approaches, views language change as the reranking of hierarchically organized UG constraints on the well-formedness of output representations (Cho 1995, Berm dez-Otero 1995). The non-serial essence of OT makes it possible to describe parallel sound shifts, whereby a number of sound changes occur in parallel, rather than sequentially. For example, in Grimm's Law, p-b shifts to psg-p, whereby the contrast is preserved, yet it is has a different segmental composition. Specifically, it is demonstrated that a comprehensive analysis of language change, and, especially, parallel sound shifts, calls for the integration of two complementary approaches within OT: faithfulness (McCarthy & Prince 1995, and others) and dispersion (Flemming 1996, Padgett 1997). In the faithfulness framework, language change is viewed as a resolution of the conflict between the tendency to save articulatory effort and the preference for the faithful mapping of input representations to their output correspondents. In the dispersion framework, language change results from the conflict between articulatory effort minimization and the preference for a maximal perceptual contrast among output forms, with no reference to inputs. In more general terms, the dispersion interaction reflects the tendency to maintain a balanced inventory, whereas a faithfulness interaction implements the preference for a transparent (i.e., non-structure changing) inventory. The integration of the two approaches amounts to claiming that each individual output form experiences two competing pressures: to be different from or similar to other output forms, and to be identical to a corresponding input form. Unlike the proponents of the dispersion-based approach (e.g., Flemming 1998), who contend that the two approaches are incompatible, I demonstrate that dispersion and faithfulness are not only compatible, but complementary in accounting for language change. Dispersion constraints exercise a stabilizing effect on the inventory. By enforcing a fixed number of contrasts, the dispersion interaction thereby delimits the range of (or censors) possible faithfulness violations. The faithfulness constraints complement the dispersion interaction, by referring to an input as the reference point relative to which the output contrast is evaluated, so that the contrast which is minimally unfaithful to the input is selected as optimal. In other words, whereas dispersion is responsible for enforcing contrast itself, faithfulness is in charge of determining the adequate segmental composition of the contrast. The analysis reveals that the integrated dispersion/faithfulness framework has advantages over earlier linguistic approaches in accounting for the sound changes which have a perceptual origin, such as Grimm's Law and Verner's Law.

This thesis explores a perspective for explaining the origins of linguistic structure that is based on considerations beyond the constraints of the language acquisition device. In contrast to the theory of Universal Grammar proposed by Chomsky, this perspective considers how the processes of language acquisition and use create a dynamical system that is capable of adapting linguistic structure to the inductive biases of learners. In this view it is possible to conceive of language adapting to aid its own survival: those languages that are more reliably and easily acquired will tend to persist for longer than their less easily learned counterparts. Thus, linguistic structures are seen as emergent, adaptive phenomena rather than preordained features of language.

The particular issue that this thesis investigates is the extent to which language adaptation can facilitate acquisition by general-purpose learners. In the Generative Grammar tradition much is made of the necessity for domain-specific constraints on the language acquisition device. (Indeed, that there must be a distinct mental com- ponent dedicated to language tasks.) This outlook is in contrast to the connectionist viewpoint, which posits far more moderately constrained, domain-general mecha- nisms. This thesis examines how language adaptation can give general-purpose, connectionist learners the appearance of being language-savvy learners.

A simulation framework is proposed in which agents attempt to communicate simple concepts to one another using sequential utterances. In earlier simulations we aim to maximise the learnability of a language for the communication task. Later simulations show how the processes of language production and acquisition, when iterated, are capable of producing such languages. In total, three series of simulations are performed.

The first series of simulations addresses the question of how linguistic structure adapts when sender and receiver disagree on the form of language that is easiest to learn. Analysis reveals that, if necessary, the structural properties of language can take on forms that compromise between the competing constraints on sender and receiver.

The second series of simulations considers the bottleneck of linguistic transmis- sion: the requirement that learners generalise from a limited set of observed utter- ances to the entire language. Results show that generalisability can be boosted in a naive, domain-general learner by allowing language to adapt to the inductive biases present in the learner.

The third and final series of simulations investigates how the dynamical charac- teristics of linguistic change depend on the properties that drive the dynamics. That is, we explore the range of conditions under which the iterated learning dynamic is suAecient to establish a learnable language throughout the population. The results of these simulations show that the iterated learning dynamic is indeed able to act as a generator of languages that general-purpose learners are capable of acquiring.

The results from these studies suggest that through the dynamics of linguistic transmission, language can adapt to the capabilities and biases of its users. Fur- thermore, that language can exploit the inductive biases of general-purpose learning mechanisms to facilitate their own acquisition, contrary to Universal Grammar's hypothesised need for an innate, domain-specific acquisition mechanism.