Eye movement analyses indicate the underlying reading strategy in the recovery of lexical readers

Abstract

Background: Psycholinguistic error analysis of dyslexic responses in various reading tasks provides the primary basis for clinically discriminating subtypes of pathological reading. Within this framework, phonology-related errors are indicative of a sequential word processing strategy, whereas lexical and semantic errors are associated with a lexical reading strategy. Despite the large number of published intervention studies, relatively little is known about changes in error distributions during recovery in dyslexic patients.

Aims: The main purpose of the present work was to extend the scope of research on the time course of recovery in readers with acquired dyslexia, using eye tracking methodology to examine word processing in real time. The guiding hypothesis was that in lexical readers a reduction of lexical errors and an emerging predominant production of phonological errors should be associated with a change to a more segmental moment-to-moment reading behaviour.

Methods & Procedures: Five patients participated in an eye movement supported reading intervention, where both lexical and segmental reading was facilitated. Reading performance was assessed before (T1) and after (T2) therapy intervention via recording of eye movements. Analyses included a novel way to examine the spatiotemporal dynamics of processing using distributions of fixation positions as different time intervals. These subdistributions reveal the gradual shifting of fixation positions during word processing, providing an adequate metric for objective classification of online reading strategies.

Outcome & Results: Therapy intervention led to improved reading accuracy in all subjects. In three of five participants, analyses revealed a restructuring in the underlying reading mechanisms from predominantly lexical to more segmental word processing. In contrast, two subjects maintained their lexical reading procedures. Importantly, the fundamental assumption that a high number of phonologically based reading errors must be associated with segmental word processing routines, while the production of lexical errors is indicative of a holistic reading strategy could not be verified.

Conclusions: Our results indicate that despite general improvements in reading performance, only some patients reorganised their word identification process. Contradictive data raise doubts on the validity of psycholinguistic error analysis as an exclusive indicator of changes in reading strategy. We suggest this traditional approach to combine with innovative eye tracking methodology in the interest of more comprehensive diagnostic strategies.

Keywords:

• Acquired dyslexia
• Aphasia
• Eye movement
• Reading strategy
• Reading treatment

The present investigation was supported by the German Science Foundation, DFG (HU 292/9-2). Two anonymous reviewers provided very helpful feedback on an earlier version of this paper. We thank all patients for participating in our study.

Notes

¹ The validity of fixation positions within words as indicators of lexical and sub-lexical processing has been confirmed in research conducted from a number of different perspectives. On a fundamental level, it appears clear that the centre of words provides the functional target for incoming initial saccades into words (Engbert & Krügel, 2010; Rayner, 1979). Looking at eye movements within words, research on the reading of complex words has indicated that saccade landing positions are sensitive to fine grain processing strategies on the orthographic and morphological level (e.g., Hyönä, Bertram, & Pollatsek, 2004; Inhoff, Radach, & Heller, 2000). Systematic relations between fixation positions and fixation durations have been documented (Radach & Heller, 2000; Vitu, McConkie, Kerr, & O’Regan, 2001), and the more sequential nature of word processing in developing readers has been shown to clearly correspond to more within-word refixations and larger oculomotor word length effects (Huestegge, Radach, Corbic, & Huestegge, 2009; Joseph, Liversedge, Blythe, White, & Rayner, 2009).

² Saccade landing sites were also plotted exclusively for trials with versus without correct responses, resulting in distributions very similar to those reported in Figure 1. This supplementary analysis suggested that there appears to be no direct relationship between real-time visuomotor behaviour and the success of word recognition as expressed in verbal responses.