(FINISHED 2017) Advancing behavioral and cognitive understanding of speech (ABACUS)


About the project

This was a European Research Council starting grant project that investigated which cognitive mechanisms allow us to use combinatorial speech. Human speech is unique because it uses a small set of basic speech sounds to make an unlimited set of possible utterances. This combinatorial structure allows us to make new words (such as “blog” or “app”) easily using speech sounds that we already know. Humans are the only apes that can do this, yet we do not know how our brains do it, nor do we know how exactly our abilities are different from those of other apes. Using novel experimental techniques to investigate human behavior and novel computational techniques to model human cognition, it was the goal of this project to find out how we deal with combinatorial speech.

The experimental part of the project investigated individual learning as well as cultural learning. Cultural learning is a recently developing experimental paradigm in which cultural evolution is simulated in the laboratory. Two different types of cultural learning were used: iterated learning, in which participants in the experiments learned from the output of other participants (thus simulating language transfer across generations) and social coordination, where a group of language users established a new communication system (thus simulating, for example, the spread of new words in a language community). Using the two types of cultural learning in conjunction with individual learning experiments makes it possible to zero in on how humans deal with combinatorial speech from three different angles. In addition, the project made a methodological contribution by comparing the strengths and weaknesses of the three different methods.

The experiments gave precise data about human behavior. Constructing computer models helped to formulate hypotheses about how our brains deal with combinatorial speech. Two main computer models were constructed in the project: one theoretical model quantifying structure in continuous signals, and one modeling an actual learning infant. A number of smaller Bayesian models were also constructed.

The project advanced cognitive, linguistic and computational science in four principal ways: it provided insight into how our unique ability for using combinatorial speech works, it taught us how this is implemented in the brain, it extended the methodology of experimental cultural learning and it created new computer models for dealing with human speech.

Research topics: