Dichotic listening (DL) is one of the most frequently used paradigms to study hemispheric asymmetry and has been employed in several neuroimaging studies. The classic behavioral DL paradigm requires the subject to give a verbal response on each trial, which may cause image artifacts due to head movements when applied to an imaging environment. In order to avoid such artifacts most studies have used modified versions of the classic DL paradigm, where no verbal response is required. The purpose of the present study was to test a new DL paradigm, specifically developed for the fMRI environment, which is based on collecting verbal responses and thus being as close as possible to the classic DL behavioral task. By employing a sparse-sampling EPI acquisition schema we attempted to limit the negative impact of overt speech on image quality. A 5-s 'silent gap' allowed for stimulus presentation and collection of a verbal response to occur between subsequent image acquisitions and served as a high-pass filter that was optimized to detect activations of interest. Hence, the contribution of response-related activations to the measured signal was reduced. Twelve healthy volunteers (six males and six females) participated in the study. In order to obtain a measure of reliability, all participants went through the classic DL paradigm three times. The results, based on the estimation of the intraclass correlation coefficient, functional probability maps as well as on laterality maps, showed consistent activation in the right and left superior temporal gyrus, left middle temporal gyrus, and right inferior temporal gyrus, thus replicating previous results with visual display and motor response DL paradigms. It is concluded that an fMRI DL paradigm based on overt verbal responses is feasible and could have general implications for future fMRI studies of speech perception and product in general.