To identify components that contribute to word-fluency performance, 126 patients referred to a Memory Clinic for suspected cognitive impairment underwent a neuropsychological examination including the Controlled Oral Word Association test (FAS; Benton & Hamsher, 1976). The number of words produced in the FAS test during six consecutive 10-s periods followed a negatively accelerated curve approaching an asymptotic level after about 30 s. The 18 FAS variables (3 letters 6 periods) were entered into an exploratory factor analysis resulting in two factors. One factor, linked to a semiautomatic and rapid retrieval from semantic memory, loaded on the initial phase of the FAS test. A second factor, linked to an effortful and slow retrieval from semantic memory, loaded on the later phase of the FAS test. Thus, two retrieval modes of semantic memory contribute differently to the initial and later part of the FAS test, which may have theoretical and practical implications.