The Trail Making Test (TMT) is a well-established test sensitive to impairment in multiple cognitive domains. There has been ambiguity about which cognitive demands are placed on the patient by TMT Part B over and above those required to perform TMT Part A. In particular, cognitive flexibility and ability to maintain a complex response set have been 2 competing hypotheses. This study preliminarily examined which of these 2 abilities primarily contributes to Part B performance. A total of 121 clinically referred Veterans Affairs patients were administered the TMT, as well as other tests of executive and other cognitive functions. Regression analyses were used to examine which tests predicted Part B performance above and beyond Part A performance. The results provide preliminary support for TMT Part B performance being more sensitive to cognitiveflexibility (operationalized as Wisconsin Card Sorting Test [WCST], percent perseverative errors) than ability to maintain set (operationalized as WCST, failure to maintain set).