A method for investigating human speech perception that combines information about human perception of speech, auditory modeling, signal detection theory, and information theory is described. For that purpose, a model for detection of signals in the auditory nerve is developed and used to analyze speech sounds. Examples are given that combine detectability information with P(heard = h |spoken = s) to access information about "cues" to some some speech sounds. These example agree qualitatively with previous results about perception of these sounds. Refinements and prospects for this approach are discussed in light of the examples.