The model presented by J. Lewis as supplemental data to the article by Palmeirim et al. (1997) is a phase model, which does not incorporate any molecular mechanism. On the other hand, the Lunatic fringe secretion model is built on a molecular mechanism. It was thus not obvious which perturbation method to use, as the perturbation magnitudes had to be similar for the comparison to be fair. The method chosen consisted in integrating the equations of the two models by small steps (20 per somitogenesis period), and at each step, deciding at random whether each individual oscillator should have its state variables updated or not (the chance of not updating was 20%). The kind of randomness in oscillator behaviour thus modelled corresponds to oscillators "lagging behind" their normal cycle for short periods of time.