Abstract
Changes in the image size of an object occur routinely and result in massive changes in neural activity at early stages of the visual system. For example, reducing the viewing distance between an observer and an object (i.e., increasing its retinal size) can produce a very different pattern of activity in early retinotopic cortical areas. However, changing image size has little consequence on our ability to recognize an object. Using fMRI adaptation in a rapid event-related design, we measured neural sensitivity to changes in object shape and image size in the human visual cortex. We show that size invariance — that is, a single neural population coding for multiple image sizes — is evident as early as dorsal retinotopic visual area V3. In ventral visual areas, size invariance is not apparent until the lateral occipital (LO) region. Both dorsal and ventral areas, however, were limited in their size invariance; a doubling of image size resulted in a signal equivalent to a change in shape indicating that large size changes are represented by a different neural population. Directing attention towards or away from the shape of the objects did not qualitatively change the pattern of results in any area examined, suggesting that limited, size-invariant representations are formed automatically by the human visual system. Finally, we address the role of perceived object size constancy on fMRI adaptation by associating changes in perceived depth with image size changes.
This work was supported by NIH NEI EY015342 (S.O.M.), NIH NEI R01 EY015261 (D.J.K.), and NGA HM1582-05-C-0003 (S.O.M. and D.J.K.).