Features and Objects
Anne Treisman · Oxford University Press eBooks · 2012
Perception seems so effortless and instantaneous, however rich and varied the visual scene may be, that it is hard to imagine the complexity of the analysis on which our experience depends. I have been interested in finding out which operations do in fact tax the system most, and which appear to take place automatically. The idea that feature integration might pose a special problem for the perceptual system dates back at least to the 1960s. Neisser (1967), following Minsky (1961), claimed that “to deal with the whole visual input at once, and make discriminations based on any combinations of features in the field, would require too large a brain to be plausible.” They suggested that the scene could first be articulated into parts, and a fixed set of pattern recognition procedures could then be applied repeatedly to each local region. In 1969 and 1973, I raised the possibility that whereas detection could be triggered by simple features, conscious awareness might depend on feature integration and that, with high attention load, errors of integration might be made (Harvey & Treisman, 1973; Treisman, 1969). Milner incorporated the same idea in his model of shape recognition (Milner, 1974). Garner’s discoveries (1970, 1974) that many perceptual dimensions are processed independently in ratings of similarity, in speeded classification tasks, and in absolute judgements made the question about how the dimensions are recombined to specify objects more cogent, as did the anatomical and physiological discoveries of many separate visual areas that appear to specialize in coding different properties (Cowey, 1979; Maunsell & Newsome, 1987; Van Essen & Maunsell, 1983; Zeki, 1981). In 1975 we began to collect data that confirmed these speculations. Attention did appear to be needed to ensure error free feature integration.