1. We compared the properties of saccades to somatosensory and visual
targets. This comparison provides insight into the translation of sens
ory signals coding target location in different sensory coordinate fra
meworks into motor commands of a common format. Vibrotactile stimuli w
ere delivered to the hands, which were fixed in position and concealed
beneath a barrier. Saccades of different directions and amplitudes we
re elicited by the same somatosensory target from different initial ey
e positions. Both monkeys and hu mans served as subjects. 2. Somatosen
sory saccades were less accurate than visual saccades in both humans a
nd monkeys. When the barrier concealing the hands was removed, somatos
ensory saccade accuracy improved. While the hands were concealed, the
visual frame of reference provided by room illumination did not greatl
y affect saccade accuracy: accuracy was not degraded in complete darkn
ess for two of three monkeys. 3. The endpoints of saccades to a single
somatosensory target varied with initial eye position for the monkeys
, but not for the human subjects. 4. We also found evidence of an effe
ct of Limb position on somatosensory saccades: when human subjects per
formed the task with crossed hands, the incidence of curved saccades i
ncreased. Saccades often began in the direction of the unstimulated ha
nd and curved markedly toward the stimulated hand. When one subject wa
s required to delay the saccade by 600-1,000 ms after target onset (th
e delayed saccade task), the saccades were straight. Somatosensory sac
cades were also straight when the hands were not crossed. 5. The react
ion times of somatosensory saccades were ion than the reaction times o
f visual saccades, and they decreased as a function of saccade amplitu
de. The delayed saccade task reduced the differences between somatosen
sory and visual saccade reaction times. The reaction times of saccades
to very dim visual targets increased into the range found for saccade
s to somatosensory targets. When the saccade target was the combinatio
n of the somatosensory and visual stimuli at the same location, the re
action time was slightly lower than for visual targets alone. 6. The p
eak velocities of somatosensory saccades were lower than those of visu
al saccades of the same amplitude. The velocities of saccades to combi
ned somatosensory and visual targets were indistinguishable from those
of saccades to visual targets alone. The differences between somatose
nsory and visual saccade velocity were maintained in the delayed trial
type. These differences suggest that the main sequence or velocity-am
plitude relationship characteristic of saccades depends on the modalit
y of the target. 7. The implications of these modality-dependent diffe
rences in accuracy, reaction time, and saccade velocity are discussed
with regard to models of the saccade generator and the coordinate tran
sformation necessary for somatosensory saccades.