Language Evolution and Computation Bibliography

It has recently been suggested that grammaticalization can be fruitfully explained by the glossogenetic mechanisms for language evolution and historical change. Contrary to this position, it is here argued that the incorporation of grammaticalization processes in the glossogenetic ontology is far from unproblematic.

Languages are far more complex than they need to be for one-to-one communication. This paper attempts to answer the question as to why that should be. The answer, it is suggested, lies in the evolution of story-telling, legend and myth as culturally-important means of expression. Myth may not mark the dawn of proto- or rudimentary language, or even the beginnings of full language, but its existence accounts at least in part for the evolution of linguistic complexity. Language co-evolved with mythology in symbolic frameworks which extended, to the limits of cognition, the capacity for verbal expression.

Present-day people derive pleasure from rhymes, rhythms and repetitive visual patterns, that is, from instances of similarity. Similarity is the basis for grouping items into categories and so setting up abstract general concepts such as ripeness or weight. In present times, such grouping by similarity is a source of pleasure; the current plethora of concepts and words denoting them derives partly from pleasure in forming them. Then the question arises: how far back in prehistory has this pleasure been a motivation? Both beads and handaxes suggest by their symmetry that hominins may have derived this pleasure-in-the-head or internal reward as far back as Acheulian time: at this time, the motivation to construct abstract general concepts and thus expand language may have already been present. The timing is open to question but the pattern of inference connecting symmetry to language is particularly direct.

Understanding language evolution in terms of cultural transmission across generations of language users raises the possibility that some of the processes that have shaped language evolution can also be observed in historical language change. In this paper, we explore how constraints on production may affect the cultural evolution of language by analyzing the emergence of the Romance languages from Latin. Specifically, we focus on the change from Latin's flexible but OV (Object-Verb) dominant word order with complex case marking to fixed SVO (Subject-Verb-Object) word order with little or no noun inflections in Romance Languages. We suggest that constraints on second language learners' ability to produce sentences may help explain this historical change. We conclude that historical data on linguistic change can provide a useful source of information relevant to investigating the cognitive constraints that affect the cultural evolution of language.

Among the many puzzling questions about language, two are salient: First, why are there any languages at all, evidently unique to the human lineage. Second, why are there so many languages? These are in fact the basic questions of origin and variation that so occupied Darwin and other evolutionary thinkers and comprise modern biologys explanatory core: why do we observe this particular array of living forms in the world and not others -- the key problem of reconciling the underlying unity of organisms with their apparent diversity, invariance and variation. Here we examine these two questions from the viewpoint of modern linguistics, biology, and dynamical system theory.

Human language is the result of a cascade of consequences from an initial mutation which provided a new ''representational'' capacity to some mirror neurons. This initial mutation has high evolvability. The mutation coincidentally allowed representations of the two substances of signs to meet in human brains, thus accounting directly for signs. Recursivity is a result of the self-organization triggered by the choatic system that emerged from this system of signs.

Humans acquire far more of their behaviour from conspecifics via culture than any other species. Our culture is larger because it accumulates, where other species' seem to stay approximately the same size (Tomasello, 1999). This chapter attempts to clarify the problem of cultural accumulation by distinguishing between the size of a culture that can be transmitted from one generation, and the extent of culture transmitted. A culture's size is determined largely by ecological constraints, and certainly homonins (and some other species) show adaptations to facilitate this. But the exponential accumulation hypothesised by (Tomasello, 1999) I claim cannot be accounted for this way, but rather is a consequence of increasing information value in semantic components. This process can be achieved through memetics --- semantics will be selected for which transmits the most information. Thus cultural evolution achieves compression of information, generating increased extent in culture even when maintaining a fixed size. I support my argument with evidence from simulations explaining the size of culture (ae and Bryson, 2007), and simulations demonstrating selection for increased extent Kirby (1999).

Language learning is an iterative process, with each learner learning from other learners. Analysis of this process of iterated learning with chains of Bayesian agents, each of whom learns from one agent and teaches the next, shows that it converges to a distribution over languages that reflects the inductive biases of the learners. However, if agents are taught by multiple members of the previous generation, who potentially speak different languages, then a single language quickly dominates the population. In this work, we consider a setting where agents learn from multiple teachers, but are allowed to learn multiple languages. We show that if agents have a sufficiently strong expectation that multiple languages are being spoken. we reproduce the effects of inductive biases on the outcome of iterated learning seen with chains of agents.

This paper questions the assumption that subject-verb (SV) structures are basic and primary and shows instead that these apparently simple structures are quite complex informationally, intonationally, semantically, and syntactically. In contrast, we point out that verb-subject (VS) structures, particularly those involving unaccusative verbs and sentence focus, are simpler and better candidates for primary structures from an evolutionary point of view. From this perspective, Agent-first (SV) structures, which have been mentioned as examples of protolinguistic ''fossils'' (e.g. Jackendoff 2002), are not as basic as previously thought.

For the last two decades, a major question for paleoanthropologists has been the origins of modernity and modern thinking. Explanations such as symbolic culture, fully syntactic language, or abstract reasoning are all too often proffered without clear or adequate operationalizations. It is purpose of the present paper to suggest both an evolutionary cognitive basis for one aspect of modern thinking and modern language, metaphors, and to offer a potential neurological substrate.In our attempt to trace the evolution of a more circumscribed component of modern cognition, we think the candidate trait should be shared, at least in part, by our closer nonhuman primates. The trait should also be evident early (ontogeny) in humans, and there should be some specifiable and demonstrable neurological substrate. Finally, there should be evidence that the trait unambiguously sets a foundation for modern thinking. We think this trait is numerosity, i.e., the ability to think about and reason with numbers.

A computational language game model is presented that shows how a population of language users can evolve from a brightness-based to a brightness+hue-based color term system. The shift is triggered by a change in the communication challenges posed by the environment, comparable to what happened in English during the Middle English period in response to the rise of dyeing and textile manufacturing c. 1150--1500. In a previous model that is able to explain such a shift, these two color categorization strategies were explicitly represented. This is not needed in our model. Instead, whether a population evolves a brightness-or a hue-based system is an emergent phenomenon that depends only on environmental factors. In this way, the model provides an explanation of how such a shift may come about without introducing additional mechanisms that would require further explanation.

This paper presents a preliminary description and analysis of prosodic features (amplitude, duration and rhythm) observed in Northern muriquis vocalizations. The northern muriqui (Brachyteles hypoxanthus) is an endangered primate species which lives in Atlantic forests of Minas Gerais and Espirito Santo, Brazil.

Individuals devote one third of their language time to mentioning unexpected events. We try to make sense of this universal behaviour within the Costly Signalling framework. By systematically using language to point to the unexpected, individuals send a signal that advertises their ability to anticipate danger. This shift in display behaviour, as compared with typical displays in primate species, may result from the use by hominins of artefacts to kill.

Language is a defining characteristic of the biological species Homo sapiens. But Chomskian Universal Grammar is not what is innate about language; Universal Grammar requires magical thinking about genes and genetics. Constraints of universal grammar are better explained in an evolutionary context by processes inherent in symbols, and by such processes as syntactic carpentry, metaphor, and grammaticalization. We present an evolutionary timeline for language, with biological evidence for the long-term evolution of the human capacity for language, and for the co-evolution of language and the brain.

There is currently considerable disagreement about the value of different theoretical accounts that have been employed to explain the evolution of communication. In this paper, we review some of the core tenets of the 'adaptationist' and the 'informational' account. We argue that the former has its strength mainly in explaining the evolution of signals and the maintenance of honest signaling, while the latter is indispensible for understanding the cognitive mechanisms underpinning signal usage, structure, and comprehension. Importantly, the informational account that incorporates linguistic concepts is necessary prerequisite to identify which design features of language are shared with nonhuman primates or other animals, and which ones constitute derived traits specific for the human lineage.

Because language doesn't fossilize, it is difficult to unambiguously time the evolutionary events leading to language in the human lineage using traditional paleontological data. Furthermore, techniques from historical linguistics are generally seen to have an insufficient time depth to tell us anything about the nature of pre- modern-human language. Thus hypotheses about early stages of language evolution have often been seen as untestable ''fairy tales''. However, the discovery of human-unique alleles, associated with different aspects of language, offers a way out of this impasse. If an allele has been subjected to powerful selection, reaching or nearing fixation, statistical techniques allow us to approximately date the timing of the selective sweep. This technique has been employed to date the selective sweep associated with FOXP2, our current best example of a gene associated with spoken language. Although the dates themselves are subject to considerable error, a series of different dates, for different language-associated genes, provides a powerful means of testing evolutionary models of language if they are explicit and span the complete time period between our separation from chimpanzees to the present. We illustrate the potential of this approach by deriving explicit timing predictions from four contrasting models of ''protolanguage.'' For example, models of musical protolanguage suggest that vocal control came early, while gestural protolanguage sees speech as a late addition. Donald's mimetic protolanguage argues that these should appear at the same time, and further suggests that this was associated with Homo erectus. Although there are too few language-associated genes currently known to resolve the issue now, recent progress in the genetic basis for dyslexia and autism offers considerable hope that a suite of such genes will soon be available, and we offer this theoretical framework both in anticipation of this time, and to spur those developing hypotheses of language evolution to make them explicit enough to be integrated within such a hypothesis-testing framework.

Slavic aspect has remained a mystery for centuries and continues to fascinate linguists. The genesis of this intricate grammar category is even a greater puzzle. This paper aims at computationally reconstructing the prerequisite for aspect -- the emergence of a system of markers for Aktionsarten. We present an experiment where artificial language users develop a conventional system as the consequence of their distributed choices in locally situated communicative acts.

This paper reviews findings on comparative primate and animal cognition. It suggests that although modern human linguistic, cognitive, and motor behaviors differ profoundly from those of great apes, primarily with respect to required mental constructional skills, an early hominin with ape-like capacities could have used non-innate, referential signals. To determine the most probable selective agents that may have motivated these first steps towards language evolution, it is necessary to look beyond the non-human primates to a wider range of animal species. When this is done, foraging adaptations emerge as the most probable selective agents for cooperative breeding and for the cognitive and behavioral suite that would eventually lead to language.

This paper adopts the category game model that simulates the coevolution of categories and their word labels to explore the effect of social structure on linguistic categorization. Instead of detailed social connections, we adopt social popularities, the probabilities with which individuals participate into language games, to denote quantitatively the general characteristics of social structures. The simulation results show that a certain degree of social scaling could accelerate the categorization process, while a much high degree of social scaling will greatly delay this process.

Cultural transmission is the primary medium of linguistic interactions. We propose an acquisition framework that involves the major forms of cultural transmissions, such as vertical, oblique and horizontal transmissions. By manipulating the ratios of these forms of transmission in the total number of transmission across generations of individuals, we analyze their roles in language evolution, based on a lexicon-syntax coevolution model. The simulation results indicate that all these forms of transmission collectively lead to the dynamic equilibrium of language evolution across generations.

When evolutionary biologists and epistemologists investigate the evolution of life, they deconstruct the problem into three research areas: they search for the units, levels and mechanisms of life's evolution. Here, it is investigated how a similar approach can be applied to evolutionary linguistics. A methodology is proposed that allows us to identify and further investigate the units, levels and mechanisms of language evolution.

Through a constructive study of grammaticalization as a potentially important process of language evolution, we have found two findings. One is that linguistic analogy, which applies linguistic rules extendedly, is a very critical for language acquisition and meaning change. The other is that inferences based on the recognition of similarity and contingency among particular meanings can realize unidirectional meaning change, a remarkable characteristic of grammaticalization. We discuss the significance of these findings in the context of the origin and the evolution of language, especially the role of linguistic analogy in creativity. Based on the discussion, a hypothetical scenario of the origin and the evolution of language is proposed.

In the early seventies, the bio-mathematician George Price developed a simple and concise mathematical description of evolutionary processes that abstracts away from the specific properties of biological evolution. In the talk I will argue argued that Price's framework is well-suited to model various aspects of the cultural evolution of language. The first part of the talk describes Price's approach in some detail. In the second part, case studies about its application to language evolution are presented.

It has been suggested that human language emerged as either a new, critical faculty to handle recursion, which linked two other existing systems in the brain, or as an exaptation of an existing mechanism, which had been used for a different purpose to that point. Of these two theories, the latter appears more parsimonious, but, somewhat surprisingly, has attracted less attention among researchers in the field. Navigation is a prime candidate for a task that may benefit from being able to handle recursion, and we give an account of the possible transition from navigation to language. In the described context, it appears plausible that the transition adding the crucial component of human language was promoted by kin selection. We show that once language is present among its speakers, it reinforces the mechanisms of kin selection, boosting such behaviour that benefit one's kin, and any such behaviour in turn boosts the use of language. The article also describes a mechanism through which language is used in lieu of kin markers to promote altruistic behaviour between potentially large communities of unrelated individuals.

Recent studies showed that three-year old children learned novel words better when the form and meaning of the words were sound symbolically related. This was the case for both children learning a language with a rich sound symbolic lexicon (Japanese) and that without (English). From this robust nature of sound symbolic facilitation, it was inferred that children's ability to use sound symbolism in word learning is the vestige of protolanguage consisting largely of sound symbolic words. We argued that sound symbolic protolanguage was able to refer to a wide range of information (not just auditory events). It had the added advantage that it was relatively easy to develop a shared open-class lexicon and it provided a stepping stone from a holophrastic protolanguage to a combinatoric protolanguage.

In this paper I consider the possibility that language is more strongly grounded in sensorimotor cognition than is normally assumed---a scenario which would be providential for language evolution theorists. I argue that the syntactic theory most compatible with this scenario, perhaps surprisingly, is generative grammar. I suggest that there may be a way of interpreting the syntactic structures posited in one theory of generative grammar (Minimalism) as descriptions of sensorimotor processing, and discuss the implications of this for models of language evolution.

In this paper we offer arguments for why modeling in the field of artificial language evolution can benefit from the use of real robots. We will propose that robotic experimental setups lead to more realistic and robust models, that real-word perception can provide the basis for richer semantics and that embodiment itself can be a driving force in language evolution. We will discuss these proposals by reviewing a variety of robotic experiments that have been carried out in our group and try to argue for the relevance of the approach.

The received view is that the first distinct word types were noun and verb (Heine \& Kuteva, 2002; Hurford, 2003a). Heine and Kuteva (2007) have suggested that the first words were noun-like entities. The present paper submits ten new arguments that support this claim. The arguments are novel implications of the reviewed evidence which is made to bear on the evolution of the linguistic predicate/argument (e.g. noun/verb) structure. The paper concludes that the evidence for noun-like entities antedating other word types is overwhelming.

This paper presents a model of lexical alignment in communication. The aim is to provide a reference model for simulating dialogs in naming game-related simulations of language evolution. We introduce a network model of alignment to shed light on the law-like dynamics of dialogs in contrast to their random counterpart. That way, the paper provides evidence on alignment to be used as reference data in building simulation models of dyadic conversations.

We examine the evolution of major grammatical forms and constructions as linguistic manifestations of human cognitive ability, based on historical data from English. We show that the complex linguistic system has arisen as more and more grammaticalized forms have accumulated. Word order and case go back to the earliest language. Tense, aspect, modality, gender, questions, negations, parataxis can be traced back to Proto-Indo-European, and they may go back further. Most crucial is the rise of embedding recursion and its product, the VO word order in Old English. This brought about the transition from the syntactic organization of the clause interwoven with discourse organization to the more strictly syntactic organization of the clause. With this transition, the periphrastic constructions of progressive, perfect and pluperfect, modal auxiliaries, periphrastic do and definite article arose due to speakers' desire to be more specific than was possible with the older forms. We also show the role of high-frequency words in the evolution of the grammatical forms.

The last decade has been a very productive one for our knowledge of our closest extinct relative, Homo neanderthalensis. A wide variety of studies has focused on various aspects of the Neandertal skeletal record and how to read it (e.g. Hublin 2009; Weaver 2009), on the chemical composition of their bones and how that might inform us on their diet (e.g. Richards and Trinkaus 2009), on their geographical distribution and their archaeological record (e.g. Roebroeks 2008) and, very importantly, on their genetic characteristics (e.g. Green et al. 2008; Briggs et al. 2009). Genetic studies indicate that modern humans and Neandertals shared a common ancestor only 500,000 to 700,000 years ago, which is also the picture emerging from studies of their physical remains (Hublin 2009). Building on the same Bauplan, two different hominin lineages emerged, in Africa the ancestors of modern humans, and in western Eurasia the Neandertals, who vanished from the record around 35,000 radiocarbon years ago. Integration of genetic data with the other lines of evidence promises to yield major breakthroughs in our understanding of the differences and similarities between these two groups of hominins in the very near future ...

The purpose of language is to encode information, so that it can be communicated. Both the producer and the comprehender of a communication want the encoding to be simple. However, they have competing concerns as well. The producer desires conciseness and the comprehender desires fidelity. This paper argues that the Minimum Description Length Principle (MDL) captures these two pressures on language. A genetic algorithm is used to evolve languages, that take the form of finite-state transducers, using MDL as a fitness metric. The languages that emerge are shown to have the ability to generalize beyond their initial training scope, suggesting that when selecting to satisfy MDL one is implicitly selecting for compositional languages.

The evolution of language capabilities is closely linked to the evolution of human brain structures. Human brain auditory cortices are anatomically and functionally asymmetrical. Studies at the microscopic level have found a thinner cortex and more widely spaced neuronal columns in the left (dominant) hemisphere, which reasonably correlate with its greater ability to discriminate speech sounds. The nature of these differences is consistent with a ''balloon model'' of brain growth, which states that as the brain white matter grows, it stretches the overlying cortex. Thus, the amount and duration of brain growth is an important factor in acquiring the ability to perceive speech. Humans have a much longer brain maturation time than any other primates (or animals). This ''extended maturation time'' allows language capabilities to evolve in the brain over time, rather than requiring them to be present at birth. The extended maturation time also must have a genetic basis, but not one specific to language, and the HAR1, G72 and FOXP2 genes might well be examples of genes which affect cortical and white matter growth. Finally, if this neuronal system can learn language without depending on specific language genes, then what could be the origin of universal grammar? Natural human grammars, like object-oriented software programs, are constrained to describe our experiential universe -- an idea mooted also by ''the early Wittgenstein'' and others. Insofar as humans mostly share the same experiential universe, our descriptions of it (our languages, some branches of mathematics) share many features; these common features can appear as a ''universal grammar.''

Segmentation and combination is a fundamental and ubiquitous feature of modern human languages. Here we explore its development in newly emergent language systems. Previous work has shown that manner and path are segmented and sequenced in the early stages of Nicaraguan Sign Language (NSL) but, interestingly, not in the gestures produced by Spanish speakers in the same community; gesturers conflate manner and path into a single unit. To explore the missing step between gesturers' conflated expressions and signers' sequenced expressions, we examined the gestures of homesigners: deaf children not exposed to a sign language who develop their own gesture systems to communicate with hearing family members. Seven Turkish child homesigners were asked to describe animated motion events. Homesigners resembled Spanish-speaking gesturers in that they often produced conflated manner+path gestures. However, the homesigners produced these conflated gestures along with a segmented manner or path gesture and, in this sense, also resembled NSL signers. A reanalysis of the original Nicaraguan data uncovered this same transitional form, primarily in the earliest form of NSL. These findings point to an intermediate stage that may bridge the transition from conflated forms that have no segmentation to sequenced forms that are fully segmented.

In this study, we tested the circumstances under which cultural evolution might lead to regularisation, even in the absence of an explicit learning bottleneck. We used an artificial language experiment to evaluate the degree of structure preservation and the extent of a bias for regularisation during learning, using languages which differed both in their initial levels of regularity and their frequency distributions. The differential reproduction of regular and irregular linguistic items, which may signal the existence of a systematicity bias, is apparent only in languages with skewed distributions: in uniformly distributed languages, reproduction fidelity is high in all cases. Regularisation does happen despite the lack of an explicit bottleneck, and is most significant in infrequent items from an otherwise highly regular language.

How can we explain the enormous amount of creativity and flexibility in spatial language use? In this paper we detail computational experiments that try to capture the essence of this puzzle. We hypothesize that flexible semantics which allow agents to conceptualize reality in many different ways are key to this issue. We will introduce our particular semantic modeling approach as well as the coupling of conceptual structures to the language system. We will justify the approach and show how these systems play together in the evolution of spatial language using humanoid robots.

Evans \& Levinson (2009) argue that language diversity is more robust than linguistic homogeneity, and also suggest that explanations for recurring patterns in language are not the product of an innate, evolved language faculty. I examine various kinds of evidence in favour of a specialized language faculty, and argue against the claim that typologically distinct languages must have distinct parsing systems.

Recent iterated language learning studies have shown that artificial languages evolve over the generations towards regularity. This trend has been explained as a reflection of the learners' biases. We test whether this learning bias for regularity is affected by culturally acquired knowledge, specifically by familiarity and literacy. The results of non-iterated learning experiments with miniature artificial musical and spoken languages suggest that familiarity helps us learn and reproduce the signals of a language, but literacy is required for regularities to be faithfully replicated. This in turn indicates that, by modifying human learning biases, literacy may play a role in the evolution of linguistic structure.

Negated sentences in Dutch child language are analyzed. It is argued that, rather than an innate UG structure, the child's acquisition procedure explains a temporary rise and fall of negative concord. It is further suggested that natural preferences of the acquisition procedure are a substantive source for grammatical universals. This evades the assumption that the evolution of the human brain as such has already produced an innate repertoire of grammatical universals.

The art of ranking things in genera and species is of no small importance and very much assists our judgment as well as our memory. You know how much it matters in botany, not to mention animals and other substances, or again moral and notional entities as some call them. Order largely depends on it, and many good authors write in such a way that their whole account could be divided and subdivided according to a procedure related to genera and species. This helps one not merely to retain things, but also to find them. And those who have laid out all sorts of notions under certain headings or categories have done something very useful.Gottfried Wilhelm von Leibniz, New Essays on Human Understanding (Leibniz, 1704)

Pronouns form a particularly interesting part-of-speech for evolutionary linguistics because their development is often lagging behind with respect to other changes in their language. Many hypotheses on pronoun evolution exist -- both for explaining their initial resilience to change as well as for why they eventually cave in to evolutionary pressures -- but so far, no one has proposed a formal model yet that operationalizes these explanations in a unified theory. This paper therefore presents a computational model of pronoun evolution in a multi-agent population; and argues that pronoun evolution can best be understood as an interplay between the level of language strategies, which are the procedures for learning, expanding and aligning particular features of language, and the level of the specific language systems that instantiate these strategies in terms of concrete words, morphemes and grammatical structures. This claim is supported by a case study on Spanish pronouns, which are currently undergoing an evolution from a case- to a referential-based system, the latter of which there exist multiple variations (which are called leismo, laismo and loismo depending on the type of change).

According to recent developments in (computational) Construction Grammar, language processing occurs through the incremental buildup of meaning and form according to constructional specifications. If the number of available constructions becomes large however, this results in a search process that quickly becomes cognitively unfeasible without the aid of additional guiding principles. One of the main mechanisms the brain recruits (in all sorts of tasks) to optimize processing efficiency is priming. Priming in turn requires a specific organisation of the constructions. Processing efficiency thus must have been one of the main evolutionary pressures driving the organisation of linguistic constructions. In this paper we show how constructions can be organized in a constructional dependency network in which constructions are linked through semantic and syntactic categories. Using Fluid Construction Grammar, we show how such a network can be learned incrementally in a usage-based fashion, and how it can be used to guide processing by priming the suitable constructions.

This paper discusses problems associated with the ''moving target argument'' (cf. Christiansen \& Chater 2008, Chater et al. 2009, see also Deacon 1997: 329, Johansson 2005:190). According to this common argument, rapid language change renders biological adaptations to language unlikely. However, studies of rapid biological evolution, varying rates of language change and recent simulations pose problems for the underlying assumptions of the argument. A critique of these assumptions leads to a richer view of language-biology co-evolution.