Dissociating Two Types of Context in Adaptive Reading
Keywords
Loading...
Authors
Issue Date
2021-01-01
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
During reading, proficient readers form contextual expectations about upcoming words
that guide eye-movements. In principle, such expectations can be based on two
very different forms of context. On the one hand, linguistic context allows readers
to predict upcoming words based purely on the preceding text. On the other hand,
parafoveal preview context allows readers to (partially) identify words before direct
fixation based on available (rightward) parafoveal information. While decades of
careful experimental research has demonstrated that readers can – in principle –
use both types of context, the relative importance of the two during natural reading
remains unclear. Here we address this issue by using computational models to quantify
both forms of contextual information on a word-by-word basis in three naturalistic
reading corpora, containing a total of 1.2 million fixations over 108 subjects. We
concurrently estimated the unique and overlapping cross-validated variance explained
by linguistic and parafoveal context, and non-contextual factors. Results demonstrated
a striking dissociation between context integration in word skipping behaviour and
reading times. On the one hand, word skipping, and therefore the decision of where to
fixate, was found to be predominantly driven by autonomous factors, linguistic and
parafoveal preview context contributed only marginally (if at all). On the other hand,
the reading time, and therefore the time required to process a word, was significantly
reduced by linguistic and parafoveal preview context. Together, these findings suggest
that readers intelligently use linguistic and parafoveal preview context to predict upcoming
words, however, these predictions only affect how long words are fixated, while
the decision of where to fixate next is driven by non-linguistic – offline – oculomotor
factors. These findings provide critical new insight into the differential utilisation of
contextual predictions in several aspects of adaptive reading.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen