In Study 1, we evaluated preference stability across 4 preference‐assessment methods for 6 individuals, 5 of whom had autism spectrum disorder and 1 of whom had traumatic brain injury. We also measured participants’ problem behavior as a corollary measure during all assessment methods. The highest mean correlation coefficients and Kendall rank coefficients of concordance across administrations were observed for the paired‐stimulus and multiple‐stimulus‐without‐replacement methods. Lower correspondence across administrations was observed for the free‐operant and response‐restriction methods. Although differentially higher levels of problem behavior did not occur with a single method, lower levels were consistently observed with the free‐operant method. During Study 2, we evaluated the implications of lower coefficients on reinforcer efficacy by comparing an initially identified and an immediately identified high‐preference stimulus in a reinforcer assessment. Initially identified and immediately identified high‐preference stimuli were equally effective reinforcers, suggesting that fluctuations in preference do not necessarily affect reinforcer efficacy in practice.