In the domain of face processing, the concept of a “face space”
has been in the literature for some time (Valentine, 1991). This
framework has been particularly influential because rather than
encoding the local features of a face, such as eyes, nose, and
mouth, it represents global patterns of individual variation. This
framework has allowed work to be done on high-level adaptation
for faces by providing a continuous, high-dimensional space. A
global scene property framework provides much of the same
function: It describes large patterns of global variation over natural
environmental categories in a continuous way, without the need to
represent the individual objects that a scene contains. As Experiments
3 and 4 also demonstrated, adaptation provides a method for
testing the utility of these candidate properties for scene tasks, such
as basic-level category recognition.
References