This study examined the occurrence of preferred EEG phase states at stimulus onset in an equiprobable auditory Go/NoGo task with a fixed interstimulus interval, and their effects on the resultant event-related potentials (ERPs). We used a sliding short-time FFT decomposition of the EEG at Cz for each trial to assess prestimulus EEG activity in the delta, theta, alpha and beta bands. We determined the phase of each 2. Hz narrow-band contributing to these four broad bands at 125. ms before each stimulus onset, and for the first time, avoided contamination from poststimulus EEG activity. This phase value was extrapolated 125. ms to obtain the phase at stimulus onset, combined into the broad-band phase, and used to sort trials into four phase groups for each of the four broad bands. For each band, ERPs were derived for each phase from the raw EEG activity at 19 sites. Data sets from each band were separately decomposed using temporal Principal Components Analyses with unrestricted VARIMAX rotation to extract N1-1, PN, P2, P3, SW and LP components. Each component was analysed as a function of EEG phase at stimulus onset in the context of a simple conceptualisation of orthogonal phase effects (cortical negativity vs. positivity, negative driving vs. positive driving, waxing vs. waning). The predicted non-random occurrence of phase-defined brain states was confirmed. The preferred states of negativity, negative driving, and waxing were each associated with more efficient stimulus processing, as reflected in amplitude differences of the components. The present results confirm the existence of preferred brain states and their impact on the efficiency of brain dynamics in perceptual and cognitive processing.