摘要：In a go/no-go procedure, pigeons were trained to discriminate a square line figure (S+) from a circle (S-). Generalization decrements to altered versions of S+ were analyzed to determine which features the pigeons attended to. In Experiment 1, the square was broken at midsegments or vertices and expanded by varying amounts in order to determine which of these potential visual features pigeons weight more heavily. Greater generalization decrements to midsegment deleted probes than to vertex deleted probes provided evidence that pigeons weight midsegments more heavily than vertices. By deleting single vertices or midsegments, selected so as to include all contour elements, Experiments 2 and 3 provided evidence that, rather than attending to one or a few areas of contour, most pigeons tend to allocate attention over most or all contour elements. In addition, random deletions produced greater generalization decrements than deletions of line segments. The results suggest that, for simple line drawings, pigeons attend to multiple contours at different locations and may integrate these contours into a representation of an object.
