Despite intriguing and suggestive clinical observations, no formal research has assessed the possible sparing of musical recognition and memory in Alzheimer's dementia (AD). A case study is presented of an 84-year old woman with severe cognitive impairment implicating AD, but for whom music recognition and memory, according to her caregivers, appeared to be spared. The hypotheses addressed were, first, that memory for familiar music may be spared in dementia, and second, that musical recognition and memory may be reliably assessed with existing tests if behavioral observation is employed to overcome the problem of verbal or written communication. Our hypotheses were stimulated by the patient EN, for whom diagnosis of AD became probable in 2000. With severe problems in memory, language, and cognition, she now has a mini-mental status score of 8 (out of 30) and is unable to understand or recall standard instructions. In order to assess her music recognition abilities, three tests from the previous literature were adapted for behavioral observation. Two tests involved the discrimination of familiar melodies from unfamiliar melodies. The third involved the detection of distortions ("wrong" notes) in familiar melodies and discrimination of distorted melodies from melodies correctly reproduced. Test melodies were presented to EN on a CD player and her responses were observed by two test administrators. EN responded to familiar melodies by singing along, usually with the words, and often continuing to sing after the stimulus had stopped. She never responded to the unfamiliar melodies. She responded to distorted melodies with facial expressions - surprise, laughter, a frown, or an exclamation, "Oh, dear!"; she never responded in this way to the undistorted melodies. Allowing these responses as indicators of detection, the results for EN were in the normal or near normal range of scores for elderly controls. As well, lyrics to familiar melodies, spoken in a conversational voice without rhythmic or pitch clues, often prompted EN to sing the tune that correctly accompanied the lyrics. EN's results provide encouraging support for our hypotheses that sparing of musical memory may be a feature of some forms of dementia and that it may be reliably and quantitatively assessed through behavioral observation. The contrast between EN's response to music and her mini-mental status is dramatic. The article concludes with several considerations why music may be preserved in dementia and suggestions to guide future research.