21 August: Limor Raviv

Social structure affects the emergence of linguistic structure: experimental evidence

Limor Raviv (Max Planck Institute for Psycholinguistics)

Tuesday 21 August 2018, 11:00-12.30, DSB 1.17

17 August: Shira Tal

Learning biases and language change: The effect of pragmatic factors on differential case marking

Shira Tal (Hebrew University of Jerusalem)

Friday 17 August 2018, 11:00-12.30, DSB 3.10

03 July: Carmen Saldana and Jenny Culbertson

Do cross-linguistic patterns of morpheme order reflect a cognitive bias? 

Carmen Saldana and Jenny Culbertson (University of Edinburgh)

Tuesday 03 July 2018, 11:30–12.30
DSB 3.10

26 June: Laura de Rooij

Cultural transmission in an iterated learning task: the effect of animated motion on language structure

Laura de Rooij (University of Edinburgh)

Tuesday 26 June 2018, 11:30–12.30
DSB 1.20
Research by Kirby, Cornish, & Smith (2008) has demonstrated that compositional structure can evolve in the laboratory over repeated transmission in an iterated learning paradigm. Reanalysis of their data by Beckner, Pierrehumbert, & Hay (2017) indicates that the semantic dimension motion lead the way in the development of compositional structure. Beckner et al. suggested this be evidence for a cognitive bias involving particular attention to motion. However, motion in Kirby et al. was represented abstractly with arrows, which therefore may have been interpreted as part of the shape, rather than actual (animated) movement. The current study investigates the emergence of compositional structure for the semantic dimensions shape, color, and animated motion, encompassing seven chains in two experiment versions. We found that not motion exerted most influence on novel language creation, but that the shape dimension did. That is, structured mappings between meanings and forms developed more strongly for shape than for color or motion. Implications of these results will be discussed.

19 June: Fausto Carcassi

The evolution of adjectival monotonicity

Fausto Carcassi (University of Edinburgh)

Tuesday 19 June 2018, 11:30–12.30
DSB 3.10
Gärdenfors (2004) presents a generalization about the meaning of nouns, namely that nominal meanings can be represented as convex extensions on conceptual spaces. Carr et al (2017) argue that convexity emerges during language evolution from competing pressures for simplicity and informativeness. We first present a new pattern in the semantic structure of gradable adjectives, which we call monotonicity. We then propose that monotonicity, like convexity, emerges as a response to pressures for simplicity and informativeness but that, unlike convexity, human pragmatic skills play a crucial role in its evolution. We develop a computational model to support this proposal (and, time allowing and considering the recent shift of CLE talks from proper talks to slightly more informal occasions for feedback, also present a yet-unrealized experimental design for feedback).

22 May: Marieke Woensdregt

A model of cultural co-evolution of language and perspective-taking

Marieke Woensdregt (University of Edinburgh)

Tuesday 22 May 2018, 11:00–12:30
G32, 7 George Square

Language relies on mindreading (a.k.a. theory of mind), as language users have to entertain and recognise communicative intentions. Mindreading abilities in turn profit from language, as language provides a means for expressing mental states explicitly, and for transmitting one’s understanding of minds to others (e.g. younger members of the population). Given this interdependence, it has been hypothesised that language and mindreading have co-evolved. I will present an agent-based model to formalise this hypothesis, which combines referential signalling with perspective-taking. In this model, agents’ communicative behaviour is probabilistically determined by an interplay between their language and their perspective on the world. In order to learn the language, learners thus have to simultaneously infer both the language and the perspective of the speaker they’re receiving input from (using Bayesian inference). Simulation results show that learners can solve this task by bootstrapping one from the other, but only if the speaker uses a language that is at least somewhat informative.
The question then becomes under what circumstances a population of these agents can evolve such an informative language from scratch. We explore two different selection pressures: a pressure for successful communication and a pressure for accurate perspective-inference. We also compare two different types of agents: literal communicators and pragmatic communicators. Pragmatic speakers optimise their communication behaviour by maximising the probability that their interlocutor will interpret their signals correctly, given their model of the interlocutor’s language and perspective. Iterated learning results show that literal agents can evolve meaningful linguistic conventions both under a pressure for communication and under a pressure on perspective-inference. The same results were found for pragmatic agents, except that pragmatic agents can achieve equally high levels of success at communicating and inferring perspectives with much more ambiguous languages, because these agents can compensate for suboptimal languages using their pragmatic ability. Time permitting, I will also present some preliminary results on the evolvability of this pragmatic ability.


15 May: Mark Steedman

Combinatory Universal Grammar

Mark Steedman (Informatics Edinburgh)

Tuesday 15 May 2018, 11:00–12:30
3.10 Dugald Stewart Building

Greenberg proposed as his 20th Universal a generalization about the possible language-specific orders over the elements of the noun-phrase (NP). Greenberg’s original statement has been modified a number of times, and a number of attempts have been made to explain its various reformulations in terms of “constraints on movement” of those elements within a single primary ordering corresponding to a universal order of merger or dominance, defined ultimately by their semantic types

The present paper begins by proposing a new generalization concerning the orderings allowed over the elements of the NP in rigid and more freely ordered languages. This generalization can be parsimonously captured in a theory of grammar without movement or other syntactic “action-at-a-distance” between non-contiguous elements. This theory predicts that only two of the twenty four permutations over these four elements are universally excluded. This prediction constitutes a formal universal, in that it follows from the theory of grammar itself, and appears to be both qualitatively and statistically confirmed by the data

The paper goes on to show that the same generalization appears to hold over a number of cases of order alternations in clausal serial-verb constructions in a number of languages.

24 April: Rob Truswell

What’s ‘that’?

Rob Truswell (University of Edinburgh)

Tuesday 24 April 2018, 11:00–12:30
G32, 7 George Square

This is a talk in two parts. The first part is a case study in syntactic change, trying to make sense of a brief window in Middle English when virtually all free relatives were introduced by “that”. This period can only be understood by a generalization of standard models of grammar change, aiming to capture the following insight: Grammars are largely composed of associations between pieces of form and pieces of meaning, so grammar change is change in those sets of associations.

The second part of the talk describes some pilot work with Richard Blythe and Simon Kirby, modelling the dynamics of change in those terms. I give some preliminary results from those models, and discuss next steps.

10 April: EvoLang dry-runs by Jon Carr and Stella Frank

Conceptual Structure Is Shaped By Competing Pressures For Simplicity And Informativeness

Jon Carr (University of Edinburgh)

Tuesday 10 April 2018, 11:00–12:30
G32 7 George Square

Languages are shaped by competing pressures from learning and communication. Learning favours simple languages, while communication favours informative ones, giving rise to the simplicity–informativeness tradeoff. Languages that evolve under this tradeoff are both maximally simple (learnable) and maximally informative (communicatively useful). This has been shown in natural language and in experimental settings. For example, Kemp and Regier (2012) showed that kinship systems exist at the optimal frontier of simplicity and informativeness. In a separate line of experimental work, Kirby, Tamariz, Cornish, and Smith (2015) showed that when artificial languages evolve under a learning pressure alone, they become simple and uninformative, and when languages evolve during communication, they become complex and informative; it is only when both pressures are at play that we find languages at the optimal frontier.

However, a recent iterated learning experiment by Carstensen, Xu, Smith, and Regier (2015) showed that artificial languages expressing spatial relationships tended to become more informative when subjected to a pressure from learning. This is a surprising result given the previous work briefly reviewed above, which says that informativeness is driven by the pressure from communication, not from learning. One potential explanation for this result lies in their measure of informativeness, communicative cost, which is sensitive to (a) the number of words that the language is comprised of (expressivity) and (b) the extent to which similar meanings are expressed by the same word (which we will term convexity). In their experiment expressivity was fixed at four words. As a result, the reduction in communicative cost they found must be due to categories evolving to become more convex, i.e. picking out increasingly tightly-clustered sets of meanings.

To demonstrate that learning favours convex categories, we conducted two experiments in which participants learned and produced a category system for stimuli varying on two dimensions, size and angle. In Experiment 1 participants were trained on one of three systems: One marking a distinction in angle, one marking a distinction in size, and one marking a distinction on both dimensions (see Fig. 1). The results indicated that the Angle-only system was easiest to learn, followed by the Size-only system; the Angle & Size system was hardest to learn, despite having the lowest communicative cost.

In Experiment 2, the output of one participant became the input to the following participant in a standard iterated learning design. An example chain is shown in Fig. 2. Over 12 generations the category system became increasingly easy to learn, as indicated by decreasing intergenerational transmission error. Furthermore, in the majority of chains, the language converged on a system marking only a distinction on the angle dimension, which participants found easiest in Experiment 1. This increase in simplicity is driven by increasing convexity.

We also found that most chains converged on fewer than four categories. This suggests that iterated learning acts as a pressure for simplicity by simultaneously decreasing expressivity and increasing convexity. However, if, as in Carstensen et al. (2015), expressivity is held constant, the learning pressure can only act through convexity: Although languages may become more informative under iterated learning, they do so not because of a pressure to be more communicatively useful, which in Carstensen et al.’s study necessarily decreases communicative cost as a side-effect of increasing convexity. This therefore suggests that, contra Carstensen et al. (2015), languages which are both simple yet informative will only emerge when pressures from learning and communication are at play. We support these conclusions with a Bayesian iterated learning model that displays strikingly similar results.


A rational model of linguistic accommodation and its potential role in language simplification

Stella Frank (University of Edinburgh)

Tuesday 10 April 2018, 11:00–12:30
G32 7 George Square

Languages with large numbers of adult learners tend to be less morphosyntactically complex than languages where adult learners are rare (Wray & Grace, attributed to deficiencies in adult language learning. Here we investigate an additional or alternative mechanism: rational accommodation by native speakers to non-native interlocutors.

Humans have a general aptitude for reasoning about the knowledge, beliefs and motivations of other individuals, including their linguistic knowledge (e.g. Clark, 1996; Ferguson, 1981). While our interlocutors’ linguistic knowledge will often be close to our own, this may not be the case in a population with many non- native speakers. We introduce a rational model of interactions between individuals capable of reasoning about the linguistic knowledge of others, and investigate the case of a non-native speaker interacting with an native speaker who reasons about their linguistic knowledge and accommodates accordingly. Our model shows that this accommodation mechanism can lead to the non-native speaker acquiring a language variant that is less complex than the original language.

We assume a simple model in which a language consists of a distribution over linguistic variants (e.g. past tense forms). Language simplification is modelled as regularisation, whereby the most frequent variant becomes more frequent; this corresponds to, and can be measured as, entropy reduction. We model the interaction between a non-native speaker and a native speaker as interaction between two rational (Bayesian) agents. Both agents have the same initial priors and update their beliefs about the language from data in the same way, but the non-native speaker has simply seen much less data. Within an interaction, the native speaker has a parametrisable tendency to accommodate to the non-native speaker: instead of simply using their own language, they use the version of the language that they believe the non-native speaker may have acquired at this stage of their learning, given limited exposure. Importantly, the native speaker does not know exactly what data the non-native has seen. Instead, the native speaker models the non- native speaker’s linguistic knowledge by integrating over possible datasets the non-native speaker might have seen.

Representative model results for a sample language are shown in Figure 1. While learners interacting with non-accommodating speakers eventually learn the original language, non-native speakers interacting with accommodating native speakers end up learning a more regular language. This is due to the combination of the limited exposure of the non-native individual, which results in highly skewed initial distributions and some probability of not having seen low-frequency variants (Hertwig, Barron, Weber, & Erev, 2004; Hahn, 2014), in conjunction with a native speaker who is aware of and accommodates this initial bias in the non- native speaker’s input, therefore providing the non-native speaker with further data which ‘locks in’ their biased starting point.

This model shows that accommodation by native speakers to non-native speakers during interaction can lead to language simplification, and therefore suggests how accommodation can explain the link between population makeup and lin- guistic complexity. The model assumes that individuals are capable of reasoning rationally about their interlocutors’ linguistic knowledge, an assumption we are currently testing empirically with human learners.

3 April: EvoLang dry-runs by Marieke Schouwstra and Jonas Nölle

Building meanings: compositionality of human language, and its evolution

Marieke Schouwstra (University of Edinburgh)

Tuesday 3 April 2018, 11:00–12:30
G32, 7 George Square

Compositionality is seen as a key feature of human language, It describes the mechanism by which complex structures and complex meanings are related to each other: the meaning of a complex expression is determined by the meanings of its constituents and the way in which they are put together. This characterisation has been a firm assumption among many linguists, but it is not the only possible view on language. I will describe the linguistic and philosophical background of the principle of compositionality, by sketching why human language is often described as compositional, and what the alternatives might be. Subsequently, I will look briefly at how cultural evolution experiments can help us answer the question how compositionality came about.

This is a talk I will give in Wendy Sandler’s workshop on Compositionality. See https://evolang.cles.umk.pl/pdf/WorkshopCompositionality.pdf

Environmental and social factors motivate the emergence of systematic categories and signs

Jonas Nölle , Marlene Staib, Riccardo Fusaroli and Kristian Tylén

Tuesday 3 April 2018, 11:00–12:30
G32, 7 George Square

While arbitrariness has long been considered a hallmark of human language, there have been increasing discussions about non-arbitrary relationships between form and meaning such as iconicity and systematicity (Dingemanse et al., 2015). We argue that these phenomena are not just two facets of non-arbitrariness, but serve orthogonal functions in the scaffolding of an efficient communication system. Iconicity is usually associated with learning and bootstrapping (Imai & Kita, is inhibited (Roberts et al., 2015; Verhoef et al., 2016) or when participants were provided with pre-established combinatorial categories (Theisen et al., 2010).

Building on this work, we tested in a series of dyadic silent gesture experiments whether systematicity is functionally adaptive and could also emerge in competition to iconicity in response to particular environmental and social factors. We hypothesized that structure and openness of the environment as well as working memory constrains (Christiansen & Chater, 2016) could affect the degree of systematicity as well as which referential features become systematized. In two experiments, participants had to communicate stimuli depicting stylized characters that afforded both iconic and systematic gestures (see Fig. 1A). Experiment 1 tested a) whether functional adaptivity of a given trait, based on its distribution in the referential environment (consisting of 14 stimuli drawn from a set of 24) and thus its discriminative value, would affect the likelihood of this trait being systematized and b) the impact of an expanding referential environment. In a 2×2 design we varied the distribution of PET vs. GENDER (7:7 vs. 10:4) and the openness of the environment (an expanding meaning space in the open vs. constant meanings across trials in the closed condition). Logistic mixed regression models (see supplementary materials) indicated that functionally more adaptive traits were indeed more likely to be systematized (p<.01) while openness of the environment had no significant effect (see Fig. 1B). Debriefing revealed that participants in the open condition frequently realized that the changing competitor images never had to be signed, which narrowed down the search space rather than enlarging it.

Experiment 2 was designed to further explore the roles of informational bottlenecks. We improved the ”openness” manipulation by increasing the number of stimuli (n=32) and referents per trial (n=16) and randomly sampling both targets and competitors for each trial in the open condition, while the set remained the same across trials in the closed condition. In addition, we added ”displacement” as a second factor. As in experiment 1, in the co-present condition, participant pairs could gesture while seeing the stimuli, whereas dyads in the displaced condition had to wait 3 seconds after the stimuli disappeared before they could gesture—effectively displacing the communicative from the referential context and simulating communication about absent entities. Our results indicate that displacement increased the tendency to systematize overall, while openness of the environment affected the temporal development of systematicity (see Fig. 1C). In the open condition systematicity kept increasing until the end of the experiment, while there was a decline in the closed condition that resembled the slope of both the closed and ”pseudo”-open condition in experiment 1, suggesting that a true open environment can override the tendency to fall back on simple iconic signs as communication becomes more efficient over time (Kirby et al., 2015).

Taken together, our results indicate that, given certain environmental affordances, systematicity can emerge in a novel communication system at the interaction level even in the presence of competing iconic solutions. In our experiment both systematic categories and signs emerged in response to their functional adaptivity to the environment and working memory constraints of the communicative situation (openness of the referential context, displacement). This could lead to variation that is amplified over cultural transmission (Kirby et al., 2015). Lastly, our study provides the first experimental semiotics study investigating the effect of displacement, a core property of language (Hockett, 1960) that has curiously been neglected in experimental studies so far.