How Do we Ignore Irrelevant Information Presented on Displays

Document Type


Publication Date



We will present empirical and computational work aiming to contribute toward a coherent theory of how humans ignore irrelevant information presented on displays. How do we ignore irrelevant information presented on displays? Ion Juvina (ijuvina@andrew.cmu.edu) Niels Taatgen (taatgen@cmu.edu) Department of Psychology, Carnegie Mellon University 5000 Forbes Avenue, Baker Hall 336A, Pittsburgh PA, 15213 USA An experiment was conducted with an isomorph of the classical Stroop task. A treatment condition included to-be-ignored extra-stimuli beside the typical Stroop stimuli. The treatment block was preceded and followed by control blocks composed of typical Stroop trials. Half of the participants received three color-patches and the other half received three words as extra stimuli. Some of the extra stimuli were randomly set to coincide with the target or distracter dimension of the main stimulus. The results show that adding to-be-ignored stimuli to the Stroop task can be disruptive or facilitative depending on the nature of these stimuli. When extra-stimuli are of the same kind as (but not identical to) the distracter dimension of the main stimulus (words) they are facilitative. When one of the extra stimuli matches (either visually or semantically) the distracter dimension of the main stimulus there is a disruptive effect on performance. These results alone would be best explained by a “lateral inhibition” account. A visual stimulus activates its own mental representation and inhibits representations of similar stimuli. However, a lateral inhibition account would also predict effects of the extra-stimuli on the target dimension of the main stimulus (color). Adding extra colors should disrupt performance on the primary Stroop task (color naming), except when one of the extra colors coincides with the target color, in which case the effect should be facilitative. None of these latter effects have been observed empirically. These results pose interesting challenges to modeling cognitive control in display-based tasks. Some modeling explorations will be presented and discussed at the Workshop.


Paper presented at the Fifteenth Annual ACT-R Workshop, Pittsburgh, PA, July 18-20, 2008.