Abstract of a poster presented at Posters on the Hill, a Council on Undergraduate Research-sponsored event, House Rayburn Office Building, Washington, DC, April 13, 2010: Wide availability of music recordings on the Internet has stimulated creation of various music recommendation systems, using information about artist, genre, etc. to help the listener find desired music. The system or search engine employed here is unusual in being based solely upon measures of melody and timbre (Manaris et al., 2008). We tested the validity of search engine choices by examining affective and physiological responses of listeners. All participants listened to a set of seven songs - an experimenter-selected original song and three similar and three dissimilar songs selected by the search engine. An additional five songs were selected for each participant based on an original song from one of the participant's three favorite genres. During listening, EEG and peripheral psychophysiological activity were continuously measured. In response to each song, participants (n = 38) rated their own activation and pleasantness, and their familiarity with and liking of the music. Participants showed significantly higher ratings of pleasantness and liking and higher heart-rate for similar than for dissimilar songs. Significantly greater frontal cortical EEG asymmetry was also found for similar than for dissimilar songs in the set of seven songs. Hierarchical linear modeling showed that familiarity and pleasantness ratings were both strong predictors of liking. Results indicate that this search engine reliably predicts affective responses to music, including their physiological components, and that liking is related to familiarity and pleasantness across diverse music selections. The results also suggest intrinsic links between human affective responses and patterns of melody and timbre.