To interpret such a stimulus competition procedure,  it is critical that the global and local information be equally discriminable. Besides the advantages outlined earlier, an additional benefit of training the pigeons with different sized stimuli was that it allowed us to select stimuli in which the discriminability (as reflected in choice accuracy) of each level was as similar as possible. Towards this end, all probe stimuli were created from sizes selected so that both levels supported equal levels of choice accuracy for a particular bird. For Birds #1H and #3N this was the medium stimulus size and for Birds #2R and #4B this was the small size, with the birds correctly discriminating the relevant letters at the global and local levels at 84% and 86.3% respectively in sessions collected just prior to the experiment. 

     Besides the adjustment for size, this equivalence was bought about in part by the birds’ continued improvement with the global-relevant conditions during the time between Experiments 2 and 3 and an additional change made prior to Experiment 3. Regardless of the individual birds’ choice mapping, we found after Experiment 2 that the letters ‘H’ and ‘N’ were often confused, especially in the global-relevant condition. Because of this, we replaced the letter ‘H’ with the block letter ‘P’ as one of the relevant letters. Thus, for this experiment the relevant letters were P, N, X, and T. After about 40 to 50 sessions in which the P replaced the H, mean accuracy with this letter across all stimulus conditions was 75.7%, and generally improved global-relevant performance relative to the other conditions.