Twenty subjects used a joystick control to track a visual target which
stepped a constant distance alternately up and down a display screen
at random intervals between 1 and 6 seconds. After 9 up and 9 down ste
ps over a period of 63 seconds, the sensitivity of the joystick/displa
y relation was covertly increased or decreased threefold. The size of
this change, being greater than twofold, meant that simply correcting
the tracking error would not be sufficient to bring the subjects quick
ly back on target. On the step following the change, all subjects init
ially overshot (if sensitivity increased) or undershot (if sensitivity
decreased) the target by a factor of about three. But within 2-4 reac
tion time intervals (i.e., before the next target step), all subjects
nevertheless reached the target. On the subsequent step all subjects a
gain initially overshot or undershot the target, but this time by a sm
aller amount than previously, indicating that partial adaptation to th
e changed sensitivity had occurred on the first step. In the subsequen
t 2-3 reaction time intervals, they again reached the target. This pat
tern was repeated over the subsequent 14 steps such that the overshoot
or undershoot decreased in a quasi-exponential fashion as adaptation
to the changed sensitivity occurred. The fact that the subjects reache
d the target regardless of the extent of sensory-motor adaptation demo
nstrates that an immediate 'compensatory' strategy began to operate wi
thin the first reaction time of detecting the changed sensory-motor re
lation. A theoretical mechanism for this strategy is proposed.