This week’s talk will be given by Kevin Stadler, who is in the 1st year of his PhD with us. “Everything you always wanted to know about modelling language change”, Tuesday 7th May, 11am, DSB 1.17. Abstract below.
“Everything you always wanted to know about modelling language change”
In this talk I will give an overview of what I’ve been up to in the first 8 months of my PhD. I will discuss the “problem” of language change, and how it’s different from the emergence of conventions/communication systems more generally. I will briefly talk about the kinds of accounts that have been put forward to explain language change, both functional and social, and then provide an overview and critique of formal models of language change. I will then show some preliminary results from a new model of momentum-based selection inspired by cultural change in other (non-linguistic) domains that spontaneously produces the s-shaped transitions observed in language change. Using this result as a starting point I will discuss some interesting hypotheses that the model relies on and predictions that it makes, all of which are up to empirical testing.
Cat will be giving a talk on her latest experiments: “The effects of communication on category structure”, Tuesday 30th April, 11am, DSB 1.17. Abstract below.
“The effects of communication on category structure”
Words divide the world into labeled categories. Languages vary in the categories they label, sometimes to the point of making cross-cutting divisions of the same space. A potential reason for this is that linguistic categories are constructed partly via communication. Previous work suggests two opposing hypotheses about how communication might contribute to category structure: 1) communication aligns individuals’ categories, creating one of a number of optimally shareable category structures (Freyd, 1983; Gärdenfors, 2000); 2) languages’ category structures diverge from an underlying shared similarity space in ways that are not necessarily optimal, but depend on the history of communicative conventions in that particular language (Malt, Sloman, & Gennari, 1999). In the current study, participants categorise images drawn from a continuous space in three conditions: 1) a whole-set non-communicative condition, where individual participants view the whole set of images and divide it into labelled categories on the basis of similarity; 2) a communicative condition, where pairs of participants create categories by playing a partnered communication game; 3) a serial non-communicative condition, where individual participants label images serially under the same memory constraints as the participants in the communicative condition. The results show that communication creates a pressure for specificity, with participants producing more categories in the communicative condition than either of the non-communicative conditions. Although equally sized categories would be most optimal for communication, both the serial non-communicative condition and the communicative condition produce more unequally sized categories than those of non-communicative participants who labeled the whole set at once. Finally, participants in both the communicative and the whole-set non-communicative conditions have more closely aligned category systems than those in the serial non-communicative condition. Taken together, the results support different aspects of each of the two hypotheses outlined above. While individuals may share basic similarity perception of a space, the memory constraints imposed by serial encounters with objects from this space lead to idiosyncratic categorisation systems that do not necessarily divide the space up optimally. However, communication works to bring these idiosyncratic non-optimal category structures into closer alignment.
A bonus Friday talk by Prof. Nikolaus Ritt, who is visiting us from Vienna. “Modeling English word stress in terms of evolutionary game theory”, Friday 26th April, 4pm, DSB 1.17. Abstract below.
“Modeling English word stress in terms of evolutionary game theory”
The paper is assumes that a theory of cultural evolution, involving the iterated transmission of knowledge constituents, might not only explain the emergence of basic structural features of human language as such, but also underlies the historically transmitted properties of specific human languages. We therefore think that also the histories of specific languages can shed light on the principles on which cultural evolution works.
On the basis of this assumption the talk tries to account for the distribution of stress patterns in the English lexicon. We focus on the problem that – at least since the Norman Conquest and the influx of a large number of Romance loan words – English word stress has been quite variable in a rather unpredictable manner. Thus, we have ˈlentil but hoˈtel, ˈhonest but roˈbust and both ˈresearchN and reˈsearchN. Contrary to extant accounts in non-evolutionary frameworks, we do not derive stress patterns from the phonotactic structures of isolated lexical items, but regard them as emerging from the co-evolution of words (with segmental and syllabic structures) on the one hand, and rhythmic patterns on the other.
The basic idea is that words will assume lexical stress patterns that produce the best rhythmic patterns most of the time when they are expressed with one another in actual utterances. We model this in terms of evolutionary game theory. The players in our model are (disyllabic) words, and their strategies are the stress patterns they may assume (i.e initial stress or final stress). In each round of the game, two words meet (in one of a set of possible utterance contexts), choose stress patterns, and receive payoffs that reflect the well-formedness of the rhythmical structure their encounter produces. In our (rather abstract and simplified) model, the ideal rhythm is assumed to be strictly alternating and both stress clashes and lapses are seen as suboptimal. Applying standard methods of evolutionary game theory, we then calculate which distribution of stress patterns will come to be evolutionarily stable in the population of English words.
As will be seen, the model makes a few surprising predictions. Among the more interesting ones is that a mixed strategy will come to be stable among disyllabic items only if the lexicon contains a critical number of monosyllables. Even more interesting (actually to the extent of being suspicious) is that the evolutionary dynamics predicted by our model seems to fit the historical development of English word stress surprisingly well.
Alan will be talking about his PhD research on “Arbitrariness of the sign revisited: An examination of the role of phonological similarity in an artificial language learning paradigm”. Tuesday 23rd April, 11am, back in our usual venue of DSB 1.17. Abstract below.
“Arbitrariness of the sign revisited: An examination of the role of phonological similarity in an artificial language learning paradigm”
Systematic mappings between forms and meanings have been shown in a number of recent studies to provide benefits for learning; isomorphisms between form and meaning allow participants to easily learn category membership, but are potentially confounding for the learning of individual tokens. Typically, the systematic form-meaning mappings that have been explored use labels where within category the phonology of the words is similar; in these cases not only are the form-meaning mappings systematic, but the languages are also substructurally systematic. In this talk I will present the results of an experiment and a connectionist model that compare these phonologically structured systematic languages to systematic languages were phonology is not isomorphic to or predictive of meaning.
Tuesday’s talk will be given by Hannah Cornish, “Systems from Sequences”, Tuesday 16th April, 11am, DSB G06. Abstract below.
In the second talk of the week, Gregory Mills (http://homepages.inf.ed.ac.uk/gmills/) will talk about “Clarifying intentions in dialogue – how miscommunication drives alignment”. Friday 12th April, 4pm, DSB 3.10, abstract below.
Gregory Mills
Clarifying intentions in dialogue – how miscommunication drives alignment
One of the most contentious debates in studies of dialogue concerns the explanatory role assigned to speakers‘ intentions. To address this issue, this talk reports a computer-mediated variant of the maze task (Pickering & Garrod, 2004; Mills and Healey 2006), which manipulates the dialogue by inserting artificial clarification requests that appear, to participants, as if they originate from each other. Two kinds of clarification were introduced: (1) Artificial “Why?” questions that query intentions associated with the plan, (2) Fragment clarification requests that that repeat a single word from the prior turn, querying the content of participants’ referring expressions. As coordination develops, “Why?” clarification requests become progressively easier to respond to, while for fragment clarification requests the converse is the case. Further, fragment clarification requests leads to interlocutors aligning quicker on more abstract and systematic referring expressions. This talk argues that this differential pattern is not arrived at via explicit negotiation of intentions, but through the tacit turn-by-turn feedback mechanisms of dialogue.
While there have long been claims about the influence of a language?s environment on its lexis and phoneme inventory, only recently have there been concentrated attempts to identify how social complexity may systematically influence the properties of linguistic structure.
In this talk, I will discuss the various claims and empirical evidence which connect communicative structure to social structure, before examining the proposed mechanisms by which different degrees of social complexity can influence grammatical characteristics. I will then present the results of my recent experimental work investigating the effect of different population structures on artificial languages, before considering the further insights provided by computer modelling of the interactions of the participants. Finally, I will look at some of the theoretical and methodological issues I have encountered during this work and the implications for research of this kind, before considering the possibilities for future study.
Fiona Jordan (University of Bristol) is giving two guest lectures next week, on “Phylogenetics for language and culture” (3pm-5pm Monday 11th March, 2pm-4m Tuesday 12th March). These lectures are primarily intended for students on the Masters in the Evolution of Language and Cognition, but all are welcome to attend – however, places are limited, so get in touch with Kenny (kenny@ling.ed.ac.uk) if you’d like to participate.
Andrea Ravignani (PhD student in the Department of Cognitive Biology at Vienna) is visiting the LEC for a few months and will do two talks while he’s here: the first one is next Tuesday, 5th March, 11am, DSB 1.17, on “The Evolution of Complex Patterning Abilities Across Language and Music”.
The Evolution of Complex Patterning Abilities Across Language and Music
Language and music are two of the most prominent traits of human beings. Direct evidence about their origins and the evolution of their neurobiological processing requirements is unattainable. A number of indirect approaches are however available. In particular, the comparative study of cognition across a wide range of animal species can elucidate whether language and music make use of cognitive skills which specifically co-evolved with these cognitive systems, as opposed to exaptations. My work focuses on pattern perception and production, investigating their homologous and analogous cognitive underpinnings in non-human animals. I show that sensitivity to dependencies in sensory stimuli, crucial for language and music processing, is present in squirrel monkeys, notably lacking vocal learning abilities. I then present a statistical approach to visual pattern perception in pigeons and keas, focussing on individual-specific heuristics used to discriminate between stimuli. After discussing the relationship between language and music, I suggest how rhythm – present in virtually all human cultures – can embody the concept of “musical syntax”. To understand whether rhythm is a uniquely human feature, I propose some critical animal experiments and hint at my ongoing contribution to the debate. Finally, I outline my future modeling work, trying to unify signaling strategies with the origins of musical rhythm.