We have recorded the eye-movements in reading of representative numbers (30+) of native speakers of English, Spanish, Chinese, Hebrew and Arabic, reading some 5000 words of equivalent newspaper text in their respective languages. Eyelink-2 technology was used to record the precise movements and fixations of the two eyes. These data were used to develop an implemented model of how the two eyes cooperate in reading to project the critical parts of the text onto the foveas of the two eyes, thus capturing a number of behaviours common to reading across different languages. This model development was itself then the input to a theoretical understanding of precisely what cognitive scientists are doing when they create and test models of cognition.
Adult readers of English, Spanish, Chinese, Hebrew and Arabic read 5000 words of newspaper text in their native languages. The movements of both of their eyes over the text were recorded as they read. These voluminous and precise data give us an unparalleled view of the same reading behaviours as they appear in languages that differ according to the direction (left-to-right or right-to-left) of the script and the relation of the script to the spoken form of the particular language.
We then interpreted these eye movements in terms of informational and anatomical constraints, and implemented the resulting theory in terms of a computational model that captured the relevant data.
We then theorized about the process of creating a theory from such data. This led us to question the conventional concept of a "universal" in a cognitive model as something that is found across all individuals. We interpreted this type of universal in terms of how theorists generalize across different instances, and we explored the inherent limitations of this type of model. We developed the neglected concept of the "concrete universal" from the established philosophical literature and applied it to our modelling of eye movements in reading. The result is a clearer understanding of what cognitive scientists do in computational cognitive modelling. We have discerned two complementary ways of applying this new approach to the architecture of a cognitive model; one way is illustrated by our modelling of eye movements in reading, the other way is illustrated by the modelling of speech perception.
We have brought together (a) classical optometric theorising, (b) new crosslinguistic behavioural data, and (c) new data concerning the visual pathways to produce an implemented computational model of binocular foveation in reading.
We have developed a parsimonious theory of the disparity between the fixations of the two eyes in reading as it manifests itself in the differing circumstances of the five languages studied.
We have reconceptualized the use of the category "universal" in cognitive modelling, deploying two types of universal – an "abstract universal" and a "concrete universal". The current and future role of cognitive modelling is clearer as a result of these distinctions.
|Effective start/end date||1/05/05 → 31/03/09|