With our growing awareness of the complexity underlying biological phenomena, our need for computational models becomes increasingly apparent. Due to their properties, biological clocks have always lent themselves to computational modelling. Their capacity to oscillate without dampening--even when deprived of all rhythmic environmental information--required the hypothesis of an endogenous oscillator. The notion of a 'clock' provided a conceptual model of this system well before the dynamics of circadian oscillators were probed by computational modelling. With growing insight into the molecular basis of circadian rhythmicity, computational models became more concrete and quantitative. Here, we review the history of modelling circadian oscillators and establish a taxonomy of the modelling world to put the large body of circadian modelling literature into context. Finally, we assess the predictive power of circadian modelling and its success in creating new hypotheses.