非表示:
キーワード:
-
要旨:
Our recognition system is sometimes highly sensitive to the viewpoint from which an object has to be identified, yet in other cases performance is view-invariant. What are the conditions under which this generalisation across view is more or less successful? One proposal is that view-sensitivity increases as the difficulty of shape discrimination increases. We report evidence from a range of studies that discrimination difficulty can both directly and indirectly influence view-sensitivity to plane misorientation and depth rotation. First, direct effects of discriminability were found in sequential picture-matching studies depicting depth-rotated views of novel, complex, 3-D objects. Shape discrimination on mismatch trials was more view-sensitive when two similar shapes were shown. Second, indirect effects of discriminability were found in similar studies but where discriminability on mismatch trials was manipulated between subjects. Match trials were identical in all mismatch contexts. View-sensitivity increased for match trials which were mixed with mismatch trials presenting more similar shapes. Third, similar indirect context effects of discriminability on view-sensitivity were found for plane-rotated views of familiar objects but these effects were weak and transient. Such effects may be hard to obtain for familiar objects where we have an expectation of the required level of shape discrimination that is difficult to alter. Nevertheless shape discriminability is clearly critical in determining the view-sensitivity of recognition. Varying mismatch trials to depict pairs of shapes which are more or less similar influences view-sensitivity directly and changing the overall context of discrimination difficulty influences view-sensitivity indirectly on identical trials.
