A convincing demonstration that a component of selective adaptation to speech is acoustic can be provided by inducing opposite direction boundary shifts on two acoustically different but phonetically identical stimulus continua. This was done by using a contingent adaptation procedure on two continua which differed only in amplitude. The “loud” stimulus continuum, which consisted of nine syllables ranging from /bow/ to /gow/, was synthesized digitally and presented to subjects at 85 dB SPL. The “quiet” continuum was made by attenuating the same syllables to 50 dB. Phoneme boundaries were measured for both continua before and after adaptation with an alternating sequence of loud /gow/'s and quiet /bow/'s. An amplitude contingent boundary shift was obtained. Adaptation shifted the “loud” continuum boundary toward /g/ and the “quiet” continuum boundary toward /b/. Since amplitude is not part of the linguistic description of speech sounds, this amplitude contingent adaptation effect must reflect changes in processing at an acoustical level of analysis.