Major neurophysiological principles of performance monitoring are not precisely known. It is a current debate in cognitive neuroscience if an error-detection neural system is involved in behavioral control and adaptation. Such a system should generate error-specific signals, but their existence is questioned by observations that correct and incorrect reactions may elicit similar neuroelectric potentials. A new approach based on a time-frequency decomposition of event-related brain potentials was applied to extract covert sub-components from the classical error-related negativity (Ne) and correct-response-related negativity (Nc) in humans. A unique error-specific sub-component from the delta (1.5-3.5 Hz) frequency band was revealed only for Ne, which was associated with error detection at the level of overall performance monitoring. A sub-component from the theta frequency band (4-8 Hz) was associated with motor response execution, but this sub-component also differentiated error from correct reactions indicating error detection at the level of movement monitoring. It is demonstrated that error-specific signals do exist in the brain. More importantly, error detection may occur in multiple functional systems operating in parallel at different levels of behavioral control.