Procedural variations affecting the reliability of the Mini-Mental State (MMS) were investigated in three studies using three different populations. Consistent results were obtained from studies of neurologically normal, at-risk, and impaired individuals which show that the test item involving serial subtractions is significantly more difficult than the alternative item involving backward spelling. MMS users are cautioned against assuming the comparability of scores across individuals, and should specify which attention and concentration task they employ in their evaluations.