Staff Member Projects

Implicit learning of abstract sequential structures, especially natural language

Fenna Poletiek

poletiekMy research focuses on implicit learning of abstract sequential structures, especially natural language. Language acquisition is an old puzzle in (psycho-) linguistics, because of the huge complexity of the language system and because of the small and noisy stimulus input available to the child for inducing the rules of language.

Consider the following sentences: "the dog the girl bites shouts", and "the girl the dog bites shouts". How can children find out the verb to which each subject noun should be attached? A new approach to the question of learnability of complex structures, comes from experimental studies using artificial grammars. I use this methodology. Within an experimental session, participants learn a grammar by mere exposure to exemplars of that grammar. By manipulating experimentally features of the grammar (more or less complex rules), and the input sample (many vs. few or long vs. short sentences), my colleagues and I try to find out determinants of this learning process. A few interesting results are that the rule learning process can be very strategic, though completely unconscious, selecting only those aspects of a structure that are useful for a current task, and neglecting unnecessary ones (Van den Bos & Poletiek, 2006), not the number of sentences determines how much we learn but the information about the grammar each conveys (Poletiek & van Schijndel, 2009) and, as recently published, the distribution of the type of sentences over time (first the short ones and only later on, the long ones) (Lai & Poletiek, 2011).

 

I am grateful to the excellent and enthusiastic contribution of Esther van den Bos (former PhD student), Jun Lai (current NWO PhD project) and many enthusiastic Master and Bachelor students, and to my colleagues at the Max Planck Institute of Psycholinguistics in Nijmegen, where I was as a guest researcher in 2011 and 2012.