Invariance of movement or stability is one aspect of skilled performance. It has generally been concluded that the more stable the movement pattern, the more efficient the motion (Brisswalter & Mottet, 1996; Holt et al. 1995). This experiment investigated the effects of freely chosen rates and experimenter paced rates on metabolic variables, Ratings of Perceived Exertion (RPE) and movement control. Eight participants with no experience of rowing were required to row on a Concept II ergometer at 100 Watts for eight 11 minute sessions; the sessions alternated between freely chosen and experimenter paced rates. Stroke rate decreased significantly with practice indicating a slowing of the stroke rate. While the reduction of stroke rate standard deviation did not reach significance there was a lowering in variability and the movement pattern became more stable as revealed by phase-plane plots of velocity against horizontal displacement. Power output variability decreased significantly with practice indicating that participants were better able to maintain the target power output of 100 W. Increased stability was accompanied by a practice-related reduction in metabolic cost. Heart Rate decreased significantly over days practice and there was a trend for oxygen consumption to decrease with practice but this was not significant. Ratings of Perceived Exertion declined significantly with practice indicating that the task was perceived as "easier" or less effortful. The effect of work at non-preferred stroke rates, contrary to expectations, was more economical than the freely chosen stroke rates. The freely chosen condition, which was more variable with respect to stroke rate, also produced significantly higher oxygen consumption and heart rate values than the non preferred condition. It is thought that the non preferred rates were more economical than the freely chosen rates because overall the stroke rates were less variable and also the stroke rate the participants were constrained to row at may have been closer to "optimal" than the freely chosen. Overall the data provide further support for the view that motor learning is associated with changes to movement control parameters that reflect increased economy of energy expenditure in achieving the task goal.