Objective: The aim was to develop an assessment procedure that was independent of language and speech production ability, to test speech feature discrimination in severe-profoundly deaf children 2 to 4 yr of age. Design: The procedure being trialed was adapted from existing procedures. The child was required to respond with a game-like motor response to a 'change' in a speech stimulus that was being presented repeatedly through a speaker. The change occurred at randomly determined times, and false alarm responses were measured during the waiting periods (while the child waited for the change). Two- to four-yr-old normally hearing children and hearing-impaired children using hearing aids and a group of 4-yr-old hearing-impaired children using cochlear implants were assessed on the task. Results: More than 82% of the 3- and 4-yr-old normally hearing and hearing-impaired children were able to complete the testing for the eight speech sound contrasts within three 20 minute sessions. Fifty percent of the 2-yr-old normally hearing and hearing- impaired children were able to condition and complete the task. All of the normally hearing children who completed the task successfully discriminated all speech sound contrasts. The performance of the hearing-impaired children using hearing aids was influenced by the degree of hearing loss and the type of speech contrast being tested. Similarly, the average performance of the children using cochlear implants was better for easier contrasts such as /ba/bi/ with contrasting vowel formant cues. Conclusions: This procedure has potential for use as a reliable clinical and research tool for assessing the development of auditory discrimination ability in 2- to 4-yr-old severe- profoundly deaf children.