Greebles actually do look like faces (but not in the way you thought).


What factors contribute to selectivity for object categories within the ventral stream? Recently, Nasr et al. and Yue et al. (2014) reported selectivity for low-level visual properties within category-selective regions. Face-selective regions were found to respond preferentially to images containing curved contours; in contrast, scene-selective regions responded preferentially to images containing rectilinear contours. These findings raise the question as to whether category-selective responses for non-face, non-place stimuli are driven by the presence or absence of such low-level properties. We were curious whether "Greebles" - non-face stimuli that, with expertise, selectively recruit the functional face network - gave rise to "face-like" responses, in part, due to their being comprised primarily of curved contours. To investigate, we developed new Greebles comprised of predominantly rectilinear contours. During fMRI subjects viewed either curved Greebles or rectilinear Greebles. Face-, place-, body part-, and object-selective regions were also localized. Subjects were not trained to identify individual Greebles, however, a 2IFC same-different matching task forced subjects to distinguish between visually dissimilar (easy) or similar (hard) Greebles. An ROI analysis of category-selective regions revealed that - in the absence of any expertise training - curved Greebles evoked higher responses in face-selective regions relative to rectilinear Greebles, while the converse was true in place-selective regions. We suggest that category selectivity arises from multiple factors: beyond experience and task demands, selectivity for low-level visual properties may help bootstrap the acquisition of ecologically-consequential visual categories or, alternatively, may arise from statistical regularities inherent in these categories (learned via other mechanisms). In either case, it appears that the original Greebles do share potentially critical visual properties with faces - it remains to be seen whether these low-level properties do influence the acquisition of visual expertise. Meeting abstract presented at VSS 2015.

DOI: 10.1167/15.12.1161

Cite this paper

@article{Shafto2015GreeblesAD, title={Greebles actually do look like faces (but not in the way you thought).}, author={Juliet P Shafto and John A. Pyles and Carol A. Jew and Michael J. Tarr}, journal={Journal of vision}, year={2015}, volume={15 12}, pages={1161} }