This paper describes an approach to speech perception testing that involves assessment of the probability of perception of phonologically significant speech pattern contrasts presented in a varying phonetic context. Such an approach provides analytic detail about a subject's access to sensory data, with minimal influence from linguistic context. Nevertheless, the results are predictive of performance at the level of conversation. For use with young children, the approach has been incorporated into a three-interval oddity test and an imitative test. Additional options under investigation include a video-game, psychoacoustic testing, and electrophysiologic testing.