Exploring and Exploiting Uncertainty in Language Processing
What we knew
There is ample evidence showing that people who are better at spotting patterns in sequences (like the ones shown in the video above) are also better at learning a language. Our statistical pattern learning ability, i.e., our ability to learn regularities in sequences of events, has been invoked to explain how knowledge of a language can emerge from exposure to usage only.
What we did not know
Is our pattern learning ability always active? If it is, does it affect our experience of language in communicative contexts generally? In other words, would individual differences in pattern learning affect readers' exploration and exploitation of language information along more dimensions that just spotting grammatical patterns?
What we did
To investigate this we collected data from monolingual Russian, healthy, educated, young adults who completed a pattern learning task (remember the video above!), and a typical everyday task of reading naturalistic sentences, word-by-word, at their own pace.
What we added
Different from other studies on statistical learning, we linked our participants’ ability to learn patterns directly with the way in which they read naturalistic sentences. We also examined the individual differences in our participants’ behaviour, focusing on specific trends of change in their behaviour throughout the experimental tasks of pattern learning and reading.
What we found
Overall we found that pattern learning ability affects the extent to which readers engage with the sentences they read. This shows how they explore language information. But the same ability also reveals what information appears to be relevant for individual readers and how they make use of it. This shows how they exploit language information.
Taken together, our findings reveal new and exciting links between human pattern learning and language cognition.
For details, read our paper Exploring and Exploiting Uncertainty: Statistical Learning Ability Affects How We Learn to Process Language Along Multiple Dimensions of Experience in Cognitive Science. It is Open Access thanks to an agreement between UoB and Wiley. Our data can be downloaded from the UoB's UBIRA and the code is in our GitHub Repository.