The predictive coding theory, although a well-supported framework for understanding brain processing, remains elusive regarding how different brain rhythms contribute to error prediction and modify the a priori probabilities of predictive events. This study addresses this issue by analyzing Event-Related Spectral Perturbations (ERSP) generated during an auditory oddball paradigm presented in both a passive and active condition. The design involved sequences of four tones, where the last tone was either predictable (standard, S), completing the scale, or less predictable (deviant, D) when the first tone was occasionally repeated. In the passive condition, participants were instructed to ignore the sounds, whereas, in the active condition, they were asked to press the up or down arrow on a keyboard depending on whether the last tone of the sequence presented a higher or lower frequency than the previous one. This experimental design aimed to bias cognitive processing towards predictable (S) or unpredictable scenarios (D) in two different conditions: passive and attentional. EEG data from 13 channels were analyzed with Morlet wavelets, revealing event-related synchronization (ERS) and desynchronization (ERD) induced by the stimuli. Early theta activity was key in computing prediction errors and updating next-trial expectations. In the active condition, theta responses were higher in D than in S trials, indicating enhanced prediction error processing with attention. Early beta activity also increased during D, likely reflecting motor adjustments. These findings emphasize the critical role of early theta rhythms and the amplifying effect of attention on prediction error processing.