This study sought to explore whether the so-called ‘paradoxical’ task-related increases in the α bandwidth of the human electroencephalogram result from increases in evoked (phase locked), as opposed to induced (non-phase locked), activity. The electroencephalograms of 18 participants were recorded while they engaged in both auditory sensory-intake tasks (listening to randomly generated ‘tunes’) and internally directed attention tasks (imagining the same randomly generated tunes) matched for auditory input. Measures of evoked (phase locked) and induced (non-phase locked) activity were compared between tasks. Increases in induced α power were found during internal attention. No experimental effects were observed for evoked activity. These results are not entirely consistent with proposals that ‘paradoxical’ α indexes the evoked inhibition of task irrelevant processing.