Same/different discrimination is a classic task for investigating relational learning in animals. Recent research suggests that pigeons can learn a trial-unique same/different discrimination, which eliminates the opportunity to memorize the training items (Brooks & Wasserman, 2008). The authors conducted three tests to elucidate the role that item-based comparison plays in this trial-unique discrimination. In the first, the authors tested the possibility that pigeons' same/different discrimination was based on textural features of the displays by creating a single, unitary texture from same and different displays; pigeons failed to discriminate these unitary textural displays. In the second, the authors varied the number of items (mosaics) in the display and the authors reproduced the characteristic decline in performance associated with fewer items. In the third, the authors systematically increased the area of two mosaics to closely match the area occupied by increasing numbers of mosaics; the results obtained with two small items persisted even when the size of the mosaics was increased. These results clearly show that pigeons' same/different discrimination was based on object-level variability and not on other properties of the displays.