The purported bidirectionality of numerosity adaptation was tested using the comparison-of-comparisons technique, which is ostensibly resistant to certain types of non-perceptual bias. Two participants (including JAS) were given these instructions: 'Adapting stimuli will be the exposed for 5 seconds in the top two quadrants. After adaptation, test stimuli will appear in all four quadrants. In each of the top two quadrants, there will be 50 dots. Select the lower quadrant whose numerosity is FARTHEST from 50. (It may be higher or lower.)' There were 100 or 25 dots in each of the two upper quadrants during adaptation. Trials were blocked by this adaptation numerosity. Bias downward, in which responses suggest the appearance of fewer than 50 dots in each upper quadrant, was significant for both participants with 100-dot adapting stimuli, confirming that this paradigm is adequate for establishing adaptation-induced perceptual biases. Neither participant's data suggested any bias (upward or downward) with 25-dot adapting stimuli.