Fig 1.
Illustration of the spatial integration of orthographic information proposed by Grainger et al. [12] and later incorporated in the architecture of OB1-reader (Snell, van Leipsig, et al. [15]).
Information about letter identities and their locations is pooled across multiple words into a single channel for location-invariant sublexical orthographic processing (via a bag-of-bigrams) and parallel word processing (bag-of-words). Relative activation levels of co-active words (illustrated by bold font) are determined by acuity, crowding, spatial attention, and length-matching. Co-active word representations compete for identification via lateral inhibitory connections. Note that this is a simplified version of the complete model that neither includes bigrams formed using interword space information (e.g., T#, #S), nor a bag of position-independent letters that would provide input to the competing word “in” (see [13]).
Fig 2.
Example of the trial procedure used in the experiments.
The question mark refers to the task that participants had to perform on the central target stimulus: lexical decision in Experiment 1 and perceptual identification in Experiment 2. It was not part of the actual procedure. Compared with the conditions that were tested in the experiment, the example stimuli are enlarged relative to screen size and are shown here in English instead of French for illustration purposes.
Fig 3.
Mean error rates (in %) as a function of flanker relatedness (unrelated vs. related) and flanker type (bigram vs. word) in Experiment 2.
Error bars indicate 95% confidence intervals.
Table 1.
Proportion of different errors made per error category in the word flanker condition calculated separately for the related and unrelated flanker conditions.
Table 2.
Proportion of different errors made per error category in the bigram flanker condition calculated separately for the related and unrelated flanker conditions.