In a bimanual four-choice reaction time (RT) task where the compatibility o
f the stimulus-response (S-R) mapping was varied, Duncan (1977) obtained a
pattern of errors suggesting that response selection was performed in two s
teps: (1) selection of the hand and (2) selection of the finger. The presen
t study is aimed at determining which of two hypotheses best accounts for D
uncan's findings. According to the first hypothesis, the two-step selection
is due to anatomical features of the to-be-executed response. In contrast,
according to the second hypothesis, such a selection reflects a more gener
al property of the response selection process. We used the same experimenta
l protocol as Duncan (1977) but with unimanual instead of bimanual response
s. The pattern of errors reported by Duncan was essentially replicated, tha
t is, subjects often selected the correct side, but failed to correctly sel
ect the finger involved in the required response. The effect of S-R compati
bility on RT was similar for the two inner fingers (middle and ring) and th
e two outer ones (fore- and little). However, this effect was larger for in
ner than for outer fingers. These results suggest a two-step response selec
tion process which allows one to discard the anatomical hypothesis.