The molecular mechanism of circadian rhythmicity is usually modeled by a transcription/translation feedback oscillator in which clock proteins negatively feed back on their own transcription to produce rhythmic levels of clock protein mRNAs, which in turn cause the production of rhythmic levels of clock proteins. This mechanism has been applied to all model organisms for which molecular data are available. This review summarizes the increasing number of anomalous observations that do not fit the standard molecular mechanism for the model organisms Acetabularia, Synechococcus, Drosophila, Neurospora, and mouse. The anomalies fall into 2 classes: observations of rhythmicity in the organism when transcription of clock genes is held constant, and rhythmicity in the organism when clock gene function is missing in knockout mutants. It is concluded that the weight of anomalies is now so large that the standard transcription/translation mechanism is no longer an adequate model for circadian oscillators. Rhythmic transcription may have other functions in the circadian system, such as participating in input and output pathways and providing robustness to the oscillations. It may be most useful to think in terms of a circadian system that uses a noncircadian oscillator consisting of metabolic feedback loops, which acquires its circadian properties from additional regulatory molecules such as the products of canonical clock genes.