The two-oscillator model of human circadian rhythmicity was analyzed when a zeitgeber relative intensity of 1, 0.5, or 0.1 was introduced into the equations. Fourier analysis was compared with dynamic analysis such as attractor reconstruction or Liapunov exponent calculation. After a 50 or 90% reduction in zeitgeber intensity, the dynamics of the system became equivalent and differed significantly from those of a system with maximal zeitgeber intensity. When 10% aleatory noise was added to the data, the analysis was still applicable, and the results obtained were essentially the same as in the absence of noise. Dynamic analysis could thus provide a distinct classification for periodic data, based on the type of analysis.