The present study used fMRI to investigate the relationship between stimulus presentation mode and attentional instruction in a free-report dichotic listening (DL) task with consonant-vowel (CV) syllables. Binaural and dichotic CV syllables were randomly presented to the subjects during four different instructional conditions: a passive listening instruction and three active instructions where subjects listened to both ears, right ear and left ear, respectively. The results showed that dichotic presentations activated areas in the superior temporal gyrus, middle and inferior frontal gyrus and the cingulate cortex to a larger extent than binaural presentations. Moreover, the results showed that increase of activation in these areas was differentially dependent on presentation mode and attentional instruction. Thus, it seems that speech perception, as studied with the DL procedure, involves a cortical network extending beyond primary speech perception areas in the brain, also including prefrontal cortex.