March 23rd: Gareth Roberts

The social modulation of learning and communicative pressures in the cultural evolution of language

Gareth Roberts, University of Pennsylvania

Tuesday, Mar 23 2021, 16:00-17:00 GMT
Zoom Details: [Please Request]

It has long been recognised that linguistic features are not distributed evenly across the world’s languages. For example, subjects are more likely to precede objects in ordinary declarative sentences, and languages with case marking are likely to have freer word order than languages without it. Such typological facts have been argued in many cases to result from pressures arising in learning and in language use. The apparent trade-off between case marking and fixed word order, for instance, has been argued to reflect a trade-off between the reduction of effort and the reduction of meaning uncertainty in learning and use. However, humans do not learn use language in a vacuum; they also inhabit social networks and use language to signal their social identity, introducing pressures that might be expected to modulate other pressures acting on language, leading in some cases to less efficient outcomes. In this talk I present a set of artificial miniature language experiments designed to investigate the role of social factors of this kind in the cultural evolution of language, presenting evidence that such factors do indeed modulate the other factors, leading in some cases to languages that are less efficient from the point of view of communication alone.

March 9th: Heidi Getz

Sentence First, Arguments After: Mechanisms of Morphosyntax Acquisition

Heidi Getz, Georgetown University

Tuesday, Mar 9 2021, 16:00-17:00 GMT
Zoom Details: [Please Request]

Natural languages contain complex grammatical patterns. For example, in German, verbs are generally morphologically finite when they occur second and morphologically non-finite when they are final, as in dein Bruder möchte in den Zoo gehen (“Your brother wants to go to the zoo”). Extensive research has documented young children’s knowledge of these morphosyntactic contingencies (Poeppel & Wexler, 1993; Deprez & Pierce, 1994), but we lack a mechanistic theory of how this knowledge is actually acquired.

One approach might be to learn the position of prosodically prominent open-class items (“verbs go 2nd or last”) and then to fill in the morphological details. Alternatively, one could work in the opposite direction, first learning the position of closed-class items (“-te goes 2nd and -en goes last”) and then fitting open-class items into the resulting structure. This second approach is counter-intuitive, but I will argue that it is the one learners take. Evidence comes from a series of miniature language experiments in which adults and children analyzed closed-class items as predictive of the presence and position of open-class items, but not the reverse. This learning asymmetry was strongest when the closed-class items had several properties typical of closed class items in natural languages (e.g., high frequency, short, no coda), suggesting that the perceptual distinctiveness of closed-class items drives this effect.

Taken together, the results suggest that early attention to closed-class items (Morgan, Meier, & Newport, 1987; Shi, Werker & Morgan, 1999; Valian & Coulson, 1988) might cause these items to serve as the constant terms in learners’ computations, allowing other patterns to be learned and represented relative to them. A learning mechanism that operates in this way would ultimately represent a broad range of language patterns in terms of the distribution of a small set of closed-class items—just as patterns are represented in modern syntactic theory (Rizzi & Cinque, 2016). The results of our experiments suggest that human languages may acquire this type of structure at least in part as a consequence of perceptual biases in the human language learner.

March 2nd: Cedric Boeckx

In defense of more gradualism in language evolution

Cedric Boeckx, ICREA/Universitat de Barcelona

Tuesday, Mar 2 2021, 11:00-12:00 GMT
Zoom Details: [Please Request]

In this talk I’d like to tackle a series of questions that, despite appearances, are related (or so I will argue). The first one revolves around the vexed question of language evolution vs language change. In discussions of this issue, the contrast between ‘evolution of the language faculty’ and ‘change from one language system to another’ is often enriched by the addition of a third situation (‘language emergence’): The change from a very simple system into a system that is “characteristically linguistic”. But how good is our characterisation of “characteristically linguistic” (or, for that matter, our characterisation of “a very simple system”)?

The second question I’ll tackle pertains to the deep history of (human) language. Typically, when this issue is raised, only two options are considered: either some key language trait is found to be shared with other hominins, or else it is said to be exclusive to our species (‘species-specific’). But what does “species-specific” mean now that we have robust evidence that the sapiens lineage separated from other hominins roughly 700k years ago, leaving perhaps as many as 500k years between then and the earliest fossils displaying a near-complete suite of what is often called “modern” traits? How about exploring a third possibility, one that takes the history of our species to have been far more interesting after the split from other hominins?

The third question has to do with the notion of self-domestication, which, thanks to the work of Okanoya, Thomas and Kirby, has been shown to have interesting explanatory implications for the evolution of specific properties of our linguistic system. Here I’ll be asking: how well do we understand the notion of (self-) domestication?; and how do we address the important challenges raised in recent years. For instance, how robust is the “domestication syndrome” and our understanding of it?

February 23rd: Yasamin Motamedi & Simon Kirby

Regularisation, naturalness, and harmony in emerging word order conventions

Yasamin Motamedi & Simon Kirby, University of Edinburgh

Tuesday, Feb 23 2021, 11:00-12:00 GMT
Zoom Details: [Please Request]

What are the factors that shape the emergence of linguistic conventions? We hypothesise that the very first stage of language evolution involves communicative strategies which are improvised on an item-by-item basis. However, as language is transmitted culturally, it begins to exhibit system-wide structural conventions.

In this talk we will use basic word order as a lens to examine the different factors that bias which conventionalised patterns are preferred. We will look at three factors, which languages exhibit to varying extents.

1. Naturalness: the preference for word orders that transparently reflect the particular semantics of what is being communicated.
2. Regularity: the preference for using the same word order predictably for a particular meaning.
3. Harmony: the preference for using the same word order across all meanings.

We will present results from three sets of online experiments which combine silent gesture with artificial language learning. These set out to answer the following questions. Is naturalness limited to improvisation? Will it be replaced by harmony through iterated learning? Is regularisation or harmonisation affected by the modality of a language?

These experiments cast light on the mechanisms involved in the very early stages of language emergence – mechanisms which lead to the system-wide structuring of conventions we see across spoken and signed languages today.

February 9th: Annie Holtz – Progress Report

Cultural evolution creates duality of patterning: the role of population structure in the simplicity/expressivity trade-off

Annie Holtz, University of Edinburgh

Tuesday, Feb 9 2021, 11:00-12:00 GMT
Zoom Details: [Please Request]

In this progress report I examine previously observed simplicity and naturalness effects in syntax and propose that these concepts can be reformulated as biases that belong to two different categories, namely system-wide and item-specific biases. I illustrate how simplicity effects, such as word order harmony, are grounded in system-wide evaluation of linguistic information by learners. Similarly, I draw on analyses of homomorphism and how the semantics of event types can condition word order as examples of how item-specific biases give rise to word order patterns. I present typological data and discuss how the effects of these two kinds of biases can combine to generate strong cross-linguistic patterns. However, the combined effect of item-specific and system-wide biases makes it hard to disambiguate their individual influence on language structure. I propose that we can disentangle the effects of these two types of biases by identifying instances in which they compete within the same linguistic structure. I also discuss how manipulating the experimental task to include varying amounts of innovation might allow us to identify under which conditions item-specific and system-wide biases influence linguistic behaviour.

February 2nd: Brian Buccola

Conceptual Alternatives/h3>
Brian Buccola, Michigan State University

Tuesday, Feb 9 2021, 11:00-12:00 GMT
Zoom Details: [Please Request]

When it comes to semantic interpretation in natural language, what we *don’t* say matters. For example, in saying “Al is a fine student”, we may convey that Al is not an outstanding student; if he were, we would have used the alternative sentence “Al is an outstanding student”. Thus, inferences arise by competition among alternative utterances. But how such inferences arise is an unsolvable problem, unless a theory of alternatives specifies what counts, among all the things that have not been said. It is usually assumed that alternatives are generated by manipulating what was actually said, e.g. by replacing one word (“fine”) with another (“outstanding”). We present a number of arguments for going beyond this powerful idea. In doing so, we argue that the level of words is not the right level of analysis for alternatives. Instead, we argue that the relation between words and alternatives is more indirect, and that alternatives are not linguistic objects in the traditional sense. Rather, we propose that competition in language is better seen as primarily determined by general reasoning preferences, or conceptual (thought) preferences (preferences that may have forged the lexicons of modern languages in the first place). We propose that such non-linguistic preferences can be measured and that these measures can be used to explain linguistic competition, non-linguistically, and more in depth.

(Joint work with Manuel Križ and Emmanuel Chemla; full paper available here:

January 26th: Simon Kirby

Cultural evolution creates duality of patterning: the role of population structure in the simplicity/expressivity trade-off

Simon Kirby, University of Edinburgh

Tuesday, Jan 26 2021, 11:00-12:00 GMT
Zoom Details: [Please Request]

One of the core design features of human language is Duality of Patterning, that is languages are organised such that meaningful signals are created out of combinations of meaningless elements (an organisation that can be seen most clearly in the existence of minimal pairs in a language’s lexicon). What are the origins of this structure, and to what extent does its explanation help us understand combinatorial behaviour more widely?

In this work-in-progress talk, I will present preliminary results from a simulation that seeks to answer this question. I will argue that combinatorial structure in signals is the result of a trade-off between a pressure for simplicity arising from learning, and a pressure for expressivity arising from communication. In this way, we can see duality of patterning as yet another aspect of language that is explained by these two forces operating together in cultural evolution.

I will also use this model to explain why there is at least one language – Al-Sayyid Bedouin Sign Language – which does not exhibit duality of patterning, and show that the trade-off between learning and communication can be affected by population structure. Specifically, the degree to which learners (as opposed to adults) contribute input to the data of the next generation of learners alters the balance of forces, promoting or inhibiting the emergence of systematic combinatorial structure.

January 19th: Noam Siegelman

Statistical learning abilities and their relation to language: Insights from individual differences

Noam Siegelman, Haskins Laboratories

Tuesday, Jan 19 2020, 16:00-17:00 GMT
Zoom Details: [Please Request]

Research on Statistical Learning (SL) had a profound impact on the study of language, highlighting aspects of the linguistic input that can be learned from experience. As a result, many researchers now hold the view that SL plays a key role in language acquisition and processing. But how can this theorized link be experimentally demonstrated? In the first part of the talk, I will review work demonstrating how the study of individual differences can reveal the underlying computations shared between SL and language. At the same time, I will highlight some challenges that this line of work faces due to the complex and multi-componential nature of SL. In light of these challenges, in the second part of the talk I will propose an alternative approach for studying the role of SL in language, which is based on an analysis of actual linguistic behavior given the statistical regularities embedded in the input. I will present studies applying this approach in the domain of reading acquisition, examining individual differences in children’s sensitivity to different types of regularities in their writing system and the relations between this variability and their emerging reading skills. I will conclude by discussing the implications of this work to SL theory, proposing a view that ties SL skills to success or failure in finding the balance between the impact of multiple regularities that co-exist in the sensory input in light of their systematicity.

December 17th: Jonas Nölle Pre-Viva Talk

How Language Adapts to the Environment: An Evolutionary Experimental Approach

Jonas Nölle, Centre for Language Evolution, University of Edinburgh

Thursday, Dec 17 2020, 14:00-14:30 GMT
Zoom Details: [Please Request]

This thesis investigates experimentally whether cross-linguistic variation in the structure of languages can be motivated by external factors, such as the social or physical environment. From observational and correlational data alone, it remains difficult to infer the exact underlying mechanisms. I present a novel experimental approach for studying the relationship between language and environment directly in the lab using referential communication games.
This talk focuses on a set of experiments looking at the complex issue of variation in spatial language that has been proposed to interact with topography (e.g., landmarks like rivers, slopes) and sociocultural factors (e.g., bilingualism, subsistence style, population density). Such factors seem to affect whether speakers rely on an egocentric or geocentric Frame of Reference (FoR), but their exact contributions remain unclear. I address this with a novel Virtual Reality (VR) paradigm that allows for an unprecedented combination of ecological validity and experimental control. In networked VR experiments, participants had to solve spatial coordination games in realistic environments such as a forest or mountain slope. Speakers of English, which is usually associated with an egocentric FoR, were less likely to use egocentric language (e.g., “the orb is to your left”) if strong environmental affordances made geocentric language more viable (e.g., “the orb is uphill from you”), indicating an effect of topography.

Follow-up experiments addressed whether the cultural ‘success’ of egocentric left/right could be motivated by its applicability across environments. For this, I combined VR with the ‘experimental semiotics’ approach, where the game is solved via a novel visual communication channel. I show how movement data from the 3D world can be correlated with invented signals to measure which FoR participants rely on. In contrast to the English data, I did not find an advantage for geocentric systems in the slope environment, and overwhelmingly egocentric systems emerged. I discuss how this could relate to task-specificity and native language background.
Finally, I show how this new way of studying spatial language with interactive VR games can be used to test hypotheses about linguistic transmission and material culture that could help explain the origins of the egocentric FoR system, which is regarded a fairly recent cultural innovation. More generally, I suggest that VR can be used to study the evolution of language in complex, multimodal settings without sacrificing experimental control.

December 1st: Julian Jara-Ettinger

The social basis of referential communication

Julian Jara-Ettinger, Yale University

Tuesday, Dec 01 2020 16:00-17:00 GMT
Zoom Details: [Please Request]

Human communication is an intrinsically social activity where we share our thoughts through sounds and movements. Accordingly, theoretical work has long argued that this capacity must rely on commonsense psychology—our ability to understand other people’s behavior in terms of unobservable mental states. Yet, classical empirical work suggests that the interaction between commonsense psychology and communication is surprisingly limited. In this talk, I will present two studies where we use computational models of communication to test the extent to which communication might rely on commonsense psychology. Our work suggests that traces of social reasoning appear even in one of the most basic forms of communication: reference. Moreover, these models diverge from, and systematically outperform, non-social communicative models that rely on an assumption of brevity in speech