Are emotional responses to music altered by musical training? Though previous studies have often found no significant difference between the emotional responses of musicians and non-musicians, the effects of training may be specific to the particular piece of music studied by the participants. This specificity hypothesis was the focus of the present study. Twenty-four singers with formal training on Javier Bustoâ€™s Gloria were compared with 24 comparably trained singers from other ensembles. Participants listened to a recording of Gloria while continuously rating their emotions in a 2-dimensional emotion space (activation on the vertical axis, pleasantness on the horizontal axis). PsySound software was later used to obtain second-by-second psychoacoustic measures of the music such as loudness, dissonance, and multiplicity (number of tones simultaneously noticed). Hierarchical linear modeling (HLM) was employed to examine relationships of participantsâ€™ emotional responses with the psychoacoustic variables as well as with attributes of the participants. Results showed that dissonance and multiplicity were both significant predictors of the two emotional response variables (dissonance: negative prediction coefficient for pleasantness, positive for activation; multiplicity: positive prediction coefficient for pleasantness, negative for activation). In addition, loudness was a positive predictor of activation. Training condition significantly affected the prediction of pleasantness by multiplicity (larger effect of multiplicity in the group without specific training). While further work is needed to determine why the effects of this particular musical feature might be altered by specific training, these findings give modest support to the hypothesis that specific musical training can measurably alter emotional responses to the pieces employed in the training.