1. The metabolic cost and the mechanical work at different speeds duri
ng uphill, level and downhill walking have been measured in four subje
cts. 2. The mechanical work has been partitioned into the internal wor
k (W(int)), due to the speed changes of body segment with respect to t
he body centre of mass (BCM), and the external work (W(ext)), related
to the position and speed changes of the BCM in the environment. 3. W(
ext) has been further divided into a positive part (W(ext)+) and a neg
ative one (W(ext)), associated with the energy increases and decreases
, respectively, over the stride period. 4. For all constant speeds the
most economical gradient has been found to be -10.2% (+/- 0.8 S.D.).
5. At each gradient there is a unique W(ext)+/W(ext)- ratio (= 1 in le
vel walking), regardless of speed, with a tendency for W(ext)- and W(e
xt)+ to vanish above +15% and below -15% gradient, respectively. 6. W(
int) is constant at each speed regardless of gradient. This is partly
explained by an only slight decrease in stride frequency at increasing
gradient. W(int) constancy implies that it has no role in determining
the optimum gradient. 7. A linear multiple regression relating W(ext)
+ and W(ext)-to the metabolic cost at different gradients showed that
negative (eff-) and positive (eff+) efficiencies decrease with increas
ing speed (from 0.912 to 0.726, and from 0.182 to 0.146, respectively)
. The eff-/eff+ ratio, however, remains rather constant (4.995 +/-0.12
5 S.D.). 8. We conclude that the measured W(ext), the W(ext)+/W(ext)-p
artitioning and eff-/eff+ ratio, i.e. the different efficiency of the
muscles used as force and brake generators, can explain the metabolic
optimum gradient at about -10%.