Language Evolution and Computation

This will be a dry run of Cat’s AMLAP talk:

Communication leads to the emergence of sub-optimal category structures

Words divide the world into labelled categories. Languages vary in the categories they label, sometimes to the point of making cross-cutting divisions of the same domain (e.g. spatial relations: Choi, McDonough, Bowerman, & Mandler, 1999). 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 dynamically creates one of a number of optimally shareable category structures (Freyd, 1983; Gärdenfors, 2000); 2) the category structures resulting from communication are not necessarily optimal, but diverge from a shared similarity space in language-specific ways (Malt, Sloman, Gennari, Shi, & Wang, 1999). We had participants categorise images drawn from a continuous space in three conditions: a) a non-communicative condition, where participants viewed the whole set of images and divided it into categories on the basis of similarity; b) a second non-communicative condition, where participants created categories by categorising the set of images in an incremental, serial fashion; c) a communicative condition, where participants created categories in the same incremental, serial fashion as condition b), but in the course of playing a partnered communication game. Communicative participants produced systems with more categories than those of non-communicative participants. Communicators’ categories were also more variably sized, dividing the space up in less balanced ways than non-communicators in condition a), with non-communicators in condition b) intermediate between the two. These results suggest that communication creates a pressure for specificity, encouraging more and finer-grained categories. In addition, supporting the second hypothesis outlined above, the negotiation process and memory constraints involved in building up a communication system from scratch lead to category systems that diverge in sub-optimal ways from an underlying similarity space.

Andrea will be giving a talk about his ongoing modelling and experimental workwork: “Who’s got Rhythm?: Understanding the Roots of Musical Timing Abilities”. Friday 14th, 4pm, DSB 3.10. Abstract below.

—–

“Who’s got Rhythm?: Understanding the Roots of Musical Timing Abilities”

Rhythm and synchronization are important concepts in musicology, linguistics, biology and physics. In this talk, I will present my ongoing work on the evolutionary basis of musical rhythm and its link to similar concepts in other disciplines. I will introduce a theoretical framework connecting human music and dance to non-human occurrences of synchronization, as in fireflies or crickets. I will present some preliminary results on pairwise synchronization in an agent-based model of chorusing, aimed at investigating rhythmic abilities in pre-musical hominids. Finally, I will describe recent pilot work exploring chimpanzees’ drumming capabilities.

Niki will be giving a talk on “Accommodation & Grammaticalization: or how syntactic structures become obligatory”, Tuesday 11th, 11am, DSB 1.17, abstract below.

—–

“Accommodation & Grammaticalization: or how synatic structures become obligatory”

This talk tries to relate the insight that grammaticalisation processes often come to embed lexical items into rigid syntactic structures with accomodation theory. Accomodation theorists like Peter Trudgill claim that linguistic innovations spread normally (if not exclusively) through face-to-face contact, in which speakers come to assimilate their usage to one another for obvious social reasons. Clearly, this raises the question who will assimilate to whom. What I would like to propose in my presentation is that the direction of accomodation may (at least not always) determined by such obvious criteria as the perceived social status or prestige of the participants, or by the perceptibility of differences, but also by the fact that the way in which participants interpret each other’s usage depends on their own linguistic competence. In the case of syntactic obligatorification, speakers for whom a specific structure is optional will perceive the usage of speakers for whom it is obligatory as differening from their own usage merely in quantitative terms and will find it easy to emulate, while speakers for whom a structure is syntactically obligatory will perceive the usage of speakers for whom it is not as deviant and may at the same time be insensitive to the (semantic/pragmatic) criteria that govern their usage of the construction. They will find it difficult to emulate it. This asymmetry with regard to mutual assimilation may contribute to explaining why grammaticalistion processes appear to be mostly unidirectional.

The actual historical development that will be used to exemplify (unfortunately not prove) the argument is the emergence of the syntactic category “definite article” from previous demonstrative pronouns.

Matt Spike (1st year PhD student) will be giving this week’s LEC talk, on “What can linguistic convention tell us about the cultural evolution of language?”. Tuesday 4th June 11am, DSB 1.17. Abstract below.

“What can linguistic convention tell us about the cultural evolution of language?”

While every human language is highly systematic, no two languages share the same system. Speech communities agree on the sounds, words and structures they use, but this occurs in the absence of any conscious planning or coordination. The question of how populations reach a consensus on the conventional aspects of language is not a trivial one. Steels (2012) claims that self-organisation and selection for communicative success alone can explain the cultural evolution of language in its entirety. In my talk, I will argue that neither of these criteria provide a satisfying explanation, but that a cultural evolutionary analysis of how conventions arise may provide some insight. I will do this by first looking at some simple conventions, both non-linguistic and linguistic. I will then compare a number of models from several different fields of the emergence of a structured convention, learned symbolic communication. Although the models make quite different assumptions about the mechanisms responsible, I will show that they all share at least certain qualities, and then discuss how this constrains the possible types of cultural evolution leading to structured conventions. Finally, I will consider the possibility that individual processes might contribute towards conventionality at several levels of organisation in language.

In a bonus LEC talk, Christian Rutz (St Andrews) will talk about his work with New Caledonian Crows. Friday 31st, 11am, DSB 1.17.

Prof. Nikolaus Ritt (Vienna), who is visiting us for 6 months, will give a masterclass on “Language change as cultural evolution”. These lectures are aimed at students on the Masters in the Evolution of Language and Cognition, but everyone is welcome to attend – places are limited by the size of the room though, so please let Kenny know if you’d like to participate, by email to kenny@ling.ed.ac.uk.

Thursday 23rd May, 2pm-3.30pm, DSB 1.17

Friday 24th May, 10am-11.30am, DSB 1.17

“Language Change as Cultural Evolution”

This course discusses (a) to what extent observable linguistic changes can be used as evidence in the attempt to understand the principles and the mechanisms that drive cultural evolution, and (b) to what extent theories of cultural evolution can help us to understand language change. Obviously, the two questions are two sides of the same epistemological coin.

We shall look at some fairly well attested developments from the history of English, involving phonological, morphological, syntactic and semantic changes. In each case we shall ask how the change can be described in evolutionary terms, i.e. as a change in the frequency of competing variants in a population of linguistic competence constituents, i.e. instructions for producing and being able to deal with specific types of linguistic behaviour. Some of the questions involved in taking this approach will turn out to be both interesting and challenging. For instance, we will be concerned with the issue of identifying the involved constituents in a way that allows one to (a) attribute them to distinct types (each token of which can be considered as a good enough copy of all others), to (b) establish in what way the types can be said to compete with each other, and to (c) count how many of them there actually are at each stage in a linguistic change. On the other hand, we will also try to identify possible (types of) factors involved in the selection of some specific constituent variant over its competitors. We will try to make a distinction between universal factors grounded in human physiology (in the widest sense), historically contingent social factors such as the relative ‘prestige’ of linguistic behaviour variants or their role in establishing group identity, as well as the selective pressures which some constituents of linguistic systems may exert on the historical stability of others.

In all discussions of specific changes, we shall discuss how potential accounts in terms of cultural evolution compare to accounts based on established, non-evolutionary theories of linguistic change.

James Winters, who is in the first year of his PhD with us, will be giving a talk: “It’s all just noise: Redundancy as a lower-bound on the emergence of systematicity”, Tuesday 14th May, 11am, DSB 1.17. Abstract below.

For this talk I will discuss the concept of redundancy and how it relates to the emergence of systematicity. First of all, I will provide a brief overview on a particular definition of redundancy as duplication of information in certain contexts, followed by a typology of the types of redundancy we observe in language. Next, I will discuss how context governs the conditions in which redundancy is promoted, and how only in conditions that favour a specific type of redundancy do we get systematicity. In particular, this relates to the role of redundancy as a robustness-enhancing device in the face of noise and unpredictability during communication. Lastly, I will discuss experimental work I’ve been doing that tests these assumptions in an Iterated Learning/Communication game set up.