The reproducibilities of various measurements of thumb abduction were compared. Two independent observers measured the thumb abduction in 30 volunteers by the following four methods: distance between the thumb tip and the flexion crease of the index finger proximal interphalangeal joint; distance between the flexion crease of the thumb interphalangeal joint and the proximal palmar crease; angle between the thumb and index metacarpals; and angle between the thumb and index proximal phalanx. Measurements were repeated in three weeks and their reproducibility was assessed by the intraclass correlation coefficient (ICC). Intra-observer reproducibilities by distances were high with ICCs between 0.74 and 0.89, compared with those by angles with ICCs between 0.28 and 0.71. Inter-observer reproducibilities by distances were also high (ICCs = 0.79 and 0.81) compared with those by angles (ICCs = 0.28 and 0.42). Assessment of thumb abduction based on distance in the first web is recommended in terms of reproducibility.