Long-range dependency has been found in most rhythmic motor signals. The origin of this property is unknown and largely debated. There is a controversy on the influence of voluntary control induced by requiring a pre-determined pace such as asking subjects to step to a metronome. We studied the cycle duration variability of 15 men pedaling on an ergometer at free pace and at an imposed pace (60 rpm). Revolution time was determined based on accelerometer signals (sample frequency 512 Hz). Revolution time variability was assessed by coefficient of variation (CV). The presence of long-range autocorrelations was based on scaling properties of the series variability (Hurst exponent) and the shape of the power spectral density (α exponent). Mean revolution time was significantly lower at freely chosen cadence, while values of CV were similar between both sessions. Long-range autocorrelations were highlighted in all series of cycling patterns. However, Hurst and α exponents were significantly lower at imposed cadence. This study demonstrates the presence of long-range autocorrelations during cycling and that voluntary intent can modulate the interdependency between consecutive cycles. Therefore, cycling may constitute a powerful paradigm to investigate the influence of central control mechanisms on the long-range interdependency characterizing rhythmic motor tasks.