Purchase this article with an account.
Dwight Kravitz, Nikolaus Kriegeskorte, Chris Baker; The pervasive influence of position on object processing: From brain to behavior. Journal of Vision 2009;9(8):816. doi: https://doi.org/10.1167/9.8.816.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Despite behavioral evidence suggesting that object recognition is position invariant (e.g. Biederman and Cooper, 1991), recent studies in humans and non-human primates have reported strong effects of position in anterior regions of the ventral visual pathway thought to be critical for object recognition. Here we show, both behaviorally and with fMRI, that object representations are tied to limited ranges of retinotopic positions.
In an event-related fMRI paradigm we presented 24 line drawings (5° from fixation) in each of the four quadrants (for a total of 96 stimulus conditions) while participants performed an orthogonal color-matching task. In separate runs, we localized two ventral stream regions thought to be critical for object recognition, Lateral Occipital (LO) and Posterior Fusiform Sulcus (PFs). We then used multivariate pattern analysis to establish the similarity of response between each pair of conditions.
Position was the primary determinant of the spatial pattern of response in both regions, with patterns always more similar within a position than between positions. These profound effects of position were also present in category selective regions for the preferred category (e.g. faces in FFA). Consistent with physiology, effects of position changes were strongest between hemifields and there was a bias for contralateral stimuli. Further, LO showed a bias for lower field stimuli, while PFs showed a bias for the upper field, consistent with their anatomical proximity to the upper and lower field representations in early visual cortex. The position effects were largely impervious to task with the same participants showing the same effects of position when scanned a second time with a categorization task.
Our behavioral results mirrored our imaging findings with greatly reduced object priming with shifts in position, particularly shifts between hemifields. We conclude that even high-level visual object representations are position-dependent.
This PDF is available to Subscribers Only