This study aimed to assess the effectiveness of randomized tracks of prerecorded cardiac sounds as a teaching tool for cardiac auscultation. The study focused on recognizing murmurs when present, distinguishing functional from organic murmurs, and detecting heart disease by auscultation. At both pre- and posttesting, 26 residents listened to 15 randomized tracks of live-recorded cardiac sounds and identified key features. The results indicate that the residents improved at detecting any murmur (66% vs 76%, p = 0.007) and functional murmur (37% vs 54%, p = 0.048), and marginally improved at detecting organic murmur (75% vs 84%, p = 0.129). Detection of absence of murmur declined slightly (69% vs 62%, p = 0.723). The posttest difference in identifying organic versus functional murmurs was striking (84% vs 54%, p < 0.001). Detection of heart disease (sensitivity) improved significantly (76% to 86%, p = 0.016), but there was scant improvement in detecting no disease (specificity) (55% vs 59%, p = 0.601). The residents increased in their ability to detect heart disease when present. However, the false-positive rate for a diagnosis of heart disease remained quite high. To ensure that appropriate referrals will be made, teaching should specifically target the confident recognition of functional murmurs.