Word recognition in natural reading with parafoveal preview

During natural reading, we do not only process the currently fixated (i.e., foveal) word, but also preprocess the upcoming, not yet fixated (i.e., parafoveal) word(s) to a certain degree. This so-called parafoveal preprocessing is the key to fluent and effective reading. Although fluent reading depends on parafoveal preprocessing, it was up to now not addressed in the domain of visual word recognition: No model of visual word recognition can account for parafoveal preprocessing during natural reading. In contrast, for most models of eye movement control during reading parafoveal preprocessing is an indispensable prerequisite. These models, however, (compared to models of visual word recognition) rely on an underspecified "dummy" engine of word recognition: Unfortunately, models of eye movement control cannot "simply" incorporate a model of visual word recognition not before such a model can account for the role of parafoveal preprocessing. This proposal has two ultimate goals: First, it aims at specifying the role of parafoveal preprocessing in visual word recognition during natural reading. By doing so, the project will prepare the ground for the establishment of future models of visual word recognition which take parafoveal preprocessing into account. Second, a potential neuronal framework is proposed in which parafoveal and foveal processing takes place simultaneously: It is proposed that parafoveal preprocessing during natural reading resembles the neural processes described as preconscious processing by Dehaene et al. (2006). Preconscious processing takes place in brain regions, which are distinct from the brain regions involved in conscious processing. This feature of the proposal accounts for the possibility that foveal and parafoveal processing occur simultaneously. For achieving the goals of the project, we will use novel methodological approaches which combine eye tracking with electrophysiological and eye tracking with functional imaging techniques. The prospect of the project is to inform future models of visual word recognition, which in turn could be incorporated into models of eye movement control during reading. This would merge two research lines which up to now formed two relatively independent sectors of research with the same scope: Reading.

The results of the project Word recognition in natural reading with parafoveal preview enhance our understanding of visual word recognition during natural reading. Specifically, our ability to preprocess information beyond the region of the current fixation (i.e., the parafoveal region) is thought to substantially contribute to our efficiency of recognizing words at first glance. Contemporary neuroimaging research on reading, however, primarily focused on the investigation of the neural correlates observed during single word presentations which constitutes a gross oversimplification of the actual performance accomplished by our brain when we read continuous text. To bridge this gap, we used concurrent recordings of eye movements and electrophysiological or haemodynamic responses which allowed us to relate natural, self-paced information processing to brain activation. Our findings showed that several effects which have been observed in single word studies generalize to natural reading. However, we found novel effects, such as task-induced differences and oculomotor effects, which would not have been observable with traditional paradigms. Moreover, transferring sophisticated paradigms from eye movement research to the realm of neuroimaging helps us to reach an understanding of which exact linguistic properties of the upcoming, parafoveal word can be extracted during foveal word recognition. First results from this project indicate that the information acquired during parafoveal preprocessing might not exceed sublexical processing which puts constraints on contemporary notions of fully lexical and semantic preprocessing of parafoveal words. Furthermore, we probed new techniques to investigate the amount of facilitation due to the presence of parafoveal information (i.e., preview benefit). We also investigated developmental aspects of parafoveal preprocessing during reading acquisition. Our results suggest a very early development of parafoveal preprocessing which, however, is tied to the phonological decoding ability of the individual child. The present project paved the way for applying ecologically valid experimental settings in neuroimaging research. Specifically, the approaches, which we pioneered in the present project, do not impose externally determined information processing (e.g., by means of predefined presentation rates) and task-demands which exceed the inherent requirements of natural reading. Thus, these novel approaches to investigate natural reading will have great potential for uncovering the neuronal underpinnings of typical and atypical (i.e., dyslexic) reading.