In this article, we are interested in the computational modeling of visual attention. We report methods commonly used to assess the performance of these kinds of models. We survey the strengths and weaknesses of common assessment methods based on diachronic eye-tracking data. We then illustrate the use of some methods to benchmark computational models of visual attention.

Abstract. This paper presents a new methodology for studying cognition, which combines eye movements (EM) and event-related potentials (ERP) to track the cognitive processes that occur during a single eye fixation. This technique, called eye-fixation-related potentials (EFRP), has the advantage of coupling accurate time measures from ERPs and the location of the eye on the stimulus, so it can be used to disentangle perceptual/attentional/cognitive factors affecting reading. We tested this new technique to describe the controversial parafoveal-on-foveal effects on reading, which concern the question of whether two consecutive words are processed in parallel or sequentially. The experiment directly addressed this question by looking at whether semantic relatedness on a target word in a reading-like situation might affect the processing of a prime word. Three pair-word conditions were tested: A semantically associated target word (horse-mare), a semantically nonassociated target word (horse-table) and a nonword (horse-twsui); EFRPs were compared for all conditions. The results revealed that early ERP components differentiated word and nonword processing within 119 ms postfixation (N1 component). Moreover, the amplitude of the right centrofrontal P140 varied as a function of word type, being larger in response to nonassociated words than to nonwords. This component might index a spatial attention shift to the target word and its visual categorization, being highly sensitive to orthographic regularity and "illformedness" of words. The P2 consecutive component (peaking at 215 ms) differentiated associated words and nonassociated words, which can account for the semantic parafoveal effect. The EFRP technique, therefore, appears to be fruitful for establishing a time-line of early cognitive processes during reading.

No abstract