In a simultaneous discrimination task, pigeons were first trained with two patterns: one rewarding (A+) and the other unrewarding (B-) that contained the same components (the symbols: c, d, ■ and <) but displayed in a different spatial layout. They were then tested for their choices of patterns: (1) A+ vs. its mirror image (MI); (2) A+ vs. its left-right reversal (LR); (3) MI vs. other layouts (OL) of the symbols; (4) LR vs. OL. In the first two conditions, A+ was chosen over its MI and LR reversal (i.e., no MI or LR confusions were found). In the last two conditions, MI and LR were not chosen over the OL, that is, they were not treated as substitutes for the A+. On the contrary, the OL stimuli were chosen over the transformations of A+. In all cases, the discriminations revealed a failure to confuse the A+ with its transformations, as predicted from work showing that the position of pattern components is important in pattern recognition.

Two experiments examined the role of spatial dimensions in pattern discrimination and judgment of similarity by pigeons. In Experiment 1, pigeons were given a symbolic matching-to-sample task in which they first learned to discriminate between two patterns (A and B) that differed in the spatial layout of an arrow inside a circle divided into four quadrants. The first training stimulus contained an arrow inside the Top Left quadrant and the tip of the arrow was pointing toward 90 degrees. The second training stimulus contained an arrow inside the Bottom Left quadrant and the arrow was pointed downwards at 180 degrees. Fourteen new patterns, consisting of all the remaining combinations of arrow orientations and arrow locations (quadrants), were then presented and their categorization by the pigeons was examined. The results showed that the two dimensions pertaining to the position of the arrow (Top/Bottom and Left/Right halves of the circle) and their interactions were more salient than the two dimensions pertaining to its orientation (Horizontal / Vertical arrows and two arrow ends). Experiment 2 showed that the position of a pattern component was encoded and used in similarity judgments even when the A and B differed along nonspatial dimensions (a rectangle vs a circle). When pigeons encountered new visual patterns and judged their similarity to old ones, they privileged the position of the pattern components over shape in their judgments.