Abstract
Although object recognition is often assumed to be position invariant or independent,the human behavioral evidence is ambiguous, with studies reporting evidence both for and against position independence. Moreover, many of these studies also suffer from potential attentional and semantic confounds. Critically, two key predictions derived from the properties of the ventral visual pathway have not been explicitly tested. First, there has been no reported test of the relative transfer between and within hemifields, despite the clear expectation from both single cell and human physiology that between hemifield translations should engage different pools of neurons and hence evidence less transfer. Second, there has been no explicit test of translations between the periphery and fovea, despite their known differences in the neural representation. Here we utilized masked subliminal priming to test the limits of position invariance in purely visual object recognition. In an object decision task, participants were asked to indicate whether a briefly presented (66ms) and masked stimulus was a whole or scrambled line drawing of an object, which could be either familiar or novel. The amount of priming observed during a later (128–256 trials) presentation of the same object served as the measure of transfer. In a series of experiments we show that 1) between hemifield translations evidence less priming than within hemifield translations, 2) priming decreased with increasing numbers of intervening stimuli, and 3) there was an asymmetry in priming between the fovea and periphery even when acuity was controlled for. These results provide evidence for only limited position invariance in object recognition.