Abstract
Intuition suggests that increasing uncertainty (the number of possible options) will make it harder to choose correctly. We compared the effects of uncertainty on human and ideal observers, where the “ideal” makes the maximum likelihood choice. In signal detection theory, efficiency is calculated as the fraction of the energy used by a human that is required by an ideal observer to attain the same performance. Humans identify a gabor or a letter in noise with an efficiency of 3% or 15%, respectively. If introducing spatial uncertainty affects the human and ideal observers differently, then there will be a change in efficiency. We assess this by comparing human vs ideal recognition of fixed-size targets (a gabor of two possible orientations or a Sloan letter of nine possible) in noise at several degrees of spatial uncertainty. Results from 38 observers show that increasing spatial uncertainty from 1 to 104 locations affects efficiency differently for the two tasks. For gabors, efficiency increased by 3.6x from 3.2%±0.5% to 11.6%±2.5%, but for letters the 5.3% efficiency was unchanged due to similar small increases in the thresholds of the human and ideal. This suggests that more complex tasks (with a greater number of more complex objects) are less affected by uncertainty.