This study was designed to develop and test an auditory event-related potential (ERP) based spelling system for a brain-computer interface (BCI) and to compare user's performance between the auditory and visual modality. The spelling system, where letters in a matrix were coded with acoustically presented numbers, was tested on a group of healthy volunteers. The results were compared with a visual spelling system. Nine of the 13 participants presented with the auditory ERP spelling system scored above a predefined criterion level control for communication. Compared to the visual spelling system, users' performance was lower and the peak latencies of the auditorily evoked ERPs were delayed. It was concluded that auditorily evoked ERPs from the majority of the users could be reliably classified. High accuracies were achieved in these users, rendering item selection with a BCI based on auditory stimulation feasible for communication.