For best experience please turn on javascript and use a modern browser!
You are using a browser that is no longer supported by Microsoft. Please upgrade your browser. The site may not present itself correctly if you continue browsing.

Afra Alishahi (Tilburg University)

Afra Alishahi presented a lecture on “What computers tell us about human language: the case of learning words”. Using computational modeling to study human language acquisition and processing is a relatively recent development, but it has become increasingly popular as an alternative research methodology. Unlike most linguistic theories, which normally deal with higher-level routines, computational modeling requires detailed specification of the properties of the input data received by the language learner and the mechanisms used for processing the data. This transparency offers many methodological advantages. When implementing a computational model, every assumption, bias or constraint about the characteristics of the input data and the learning mechanism has to be specified. Moreover, unlike experimental studies with human subjects, the researcher has full control over all input data that the model receives in its lifetime. This allows for precise analysis of the input’s effect on the behaviour of the model. Also, the performance of different mechanisms on the same data set can be compared, something that is almost impossible to achieve in experimental studies on humans. Finally, models can make explicit predictions about the outcome of language tasks based on specific input, and under previously unstudied conditions.

As an example of using computational modeling for studying language learning, Alishahi presented a model of cross-situational word learning in children. The model adopts a probabilistic mechanism for detecting the meaning elements that are common across all the situational contexts of a word usage. It successfully learns word meanings as probabilistic associations between words and semantic elements and draws only on general cognitive abilities. The simulation results of our model demonstrate that much about word meanings can be learned from naturally occurring child-directed data, while making novel predictions about understudied learning scenarios.


Mark Steedman (University of Edinburgh)

Mark Steedman gave a lecture on the topic of Using Linguistic Knowledge in Natural Language Processing. Not too many years ago, Linguists, Psycholinguists, and Computational Linguists used to work together, using a common theoretical base with the common goal of understanding and replicating human language understanding.

In recent years, researchers in these areas have diverged, with many linguists denying the relevance of performance and even semantics to their study, while psychologists and computationalists complain that available linguistic theories are impractical for processing tasks of the size and degree of incrementality that they have to deal with.

This talk argued that it has become necessary for these areas to get back together. The study of performance in application to human sentence processing and tasks like machine translation will not make further progress without corpora annotated with representations that are much closer to the conceptual representations underlying all human languages than any that currently exist. Conversely, theoretical linguistics needs models of language acquisition that explain how the apparent diversity of attested languages can be semantically supported by such a representation and acquired from it by children.

The talk argued that some recent discoveries in theoretical linguistics make it possible to begin drawing up a specification of a deep semantic representation that will fulfil both of these requirements, and that recent discoveries in algorithmic natural language processing can be used to model children’s acquisition of their native language from ambiguous, noisy, and incomplete meaning representations, paired with sentences. The talk concluded by considering some implications of this research program for the future of Machine Translation.


Marc Slors (Radboud University)

Marc Slors gave a SMART lecture in December. The title of the talk was “Folk-Psychology as Reconstruction: Why Social-Cognitive Processes Need Not Resemble Their Linguistic Representations”. In his talk, he argued that folk-psychological descriptions of others are often reconstructions of implicit, non-reflective assessments of others. These reconstructions serve social purposes such as e.g. offering excuses (“I’m sorry, I thought you wanted me to… etc.”). The reflective use of folk-psychology as a social cognitive tool is an infrequently used ability that is derived from this reconstructive use.

This view does not logically rule out the possibility that implicit assessment of others involves a non-conscious version of our explicit folk-psychology: a sub-personal theory of mind. However, Marc Slors argued that there are good reasons not to model our implicit social cognitive abilities on the folk-psychological reconstructions we give of them.


Ray Jackendoff (Tufts University)

In February, Ray Jackendoff presented a lecture on Language, Meaning, and Rational Thought.


Lauren Stewart (Goldsmiths, University of London)

Lauren Stewart, Reader in Psychology and founding director of the MSc in Music, Mind and Brain at the University of London gave a lecture on Congenital Amusia. The ability to make sense of musical sound has been observed in every culture since the beginning of recorded history. In early infancy, it allows us to respond to the sing-song interactions from a primary caregiver and to engage in musical play. In later life it shapes our social and cultural identities and modulates our affective and emotional states. But a few percent of the population fail to develop the ability to make sense of or engage with music. Individuals with congenital amusia cannot recognize familiar tunes, cannot tell one tune from another, frequently complain that music sounds like a “din” and avoid the many social situations in which music plays a role. In this talk, Lauren Stewart presented data from perceptual experiments suggesting that individuals with amusia are insensitive to pitch direction and are unable to retain pitch information in memory. In addition, she discussed ongoing genetic and neuroimaging approaches that we are using to characterize this disorder. The study of disordered musical development sets in sharp relief the perceptual and cognitive abilities which most of us take for granted and give us a unique chance to investigate how musical perceptual ability develops, from the level of the gene to the brain development and the emergence of a complex and fundamental human behaviour. 


SMART CS Debate on Language, thought & recursion

SMART Cognitive Science Debate on Language, Recursion and Cognition with: Daniel Everett (Bentley University), Rens Bod (ILLC), Hedde Zeijlstra (Goettingen), Kees Hengeveld (ACLC).


Eve Clark (Stanford University)

In March, Eve Clark gave a talk on: Language, Interaction, and Acquisition. In her talk, she proposed that children both learn a first language and learn how to use it in interaction.  Interaction allows children to discover the forms to use and gives them feedback on those forms for conveying particular meanings, but without disrupting the ongoing exchange.  She gave evidence for the role of interaction in the process of acquisition from (a) parental reformulations, (b) offers of new words, and (c) offers of information about new-word meanings.


Michel DeGraff (MIT)

The June SMART Lecture was presented by: Michel DeGraff, Associate Professor of Linguistics at Massachusetts Institute of Technology. Title: A hitchhiker’s guide to Cognitive Science via Creole studies. In this talk DeGraff used insights from Creole study as a starting point to investigate larger issues in language acquisition and language change—and in cognitive science more generally.  In so doing, he elaborated the bases for a “Null Theory of Creole Formation,” which is based on the interaction between second- and first-language acquisition in contact situations.   Such an interaction seems relevant to, not only Creole formation, but to all cases of language change in the context of language contact.  Indeed, such a theory makes no sui generis stipulation about any exceptional “Creole typology” or any creolization-specific diachronic processes.  On the contrary, his proposed framework treats, in uniformitarian fashion, various properties of Creole languages, including properties that are derived from the superstrate or substrate languages, or some combination of both, with various sorts of simplification AND complexity-inducing innovations.  The structural patterns underlying these innovations seem germane to other instances of language change in the scope of the Comparative Method in historical linguistics.  In his perspective, Atlantic Creoles are all genealogically related to their Germanic or Romance ancestors, once the Comparative Method is duly applied.

The “Null Theory” was contrasted with approaches whereby Creole languages are either: (i) excluded from the scope of the Comparative Method  (Thomason & Kaufman) or (ii) claimed to constitute an exceptional and simplest typology (McWhorter; Parkvall; Bakker, Daval-Markussen, Parkvall & Plag; etc.). Such a contrast helps evaluate, and improve upon, recent applications of computational phylogenetics in historical linguistics and Creole studies.