Meyerholz, Irzinger, Withöft, Gerlach, and Pohl (2019) reported on a comparably large effect (d = 1.21) of a contingent biofeedback procedure on cardiac accuracy as assessed by the heartbeat tracking task. However, this task has recently been criticized as a measure of interoceptive accuracy. We aimed to replicate this finding by using the well-validated heartbeat discrimination task and to compare the biofeedback with a deep breathing and a control condition (viewing a film clip). The trial was preregistered at open science framework (https://osf.io/9fxn6). Overall, 93 participants were randomized to one of the three conditions and the heartbeat discrimination task was presented prior and after the 20-minutes training sessions. The study had a power of .86 to detect a medium-sized effect in the biofeedback group and a power of .96 to detect a medium-sized interaction of intervention group and time. A general tendency for improvement in heartbeat detection accuracy was found across intervention groups (d = 0.19, p = .08); however, groups did not differ significantly. In particular, there was no significant interaction of intervention group and time (f = .00, p = .98) and no reliable effect for the biofeedback group (d = 0.15, p = .42). One limitation is that a different, but well-validated task was used to quantify interoceptive accuracy. This study suggests that biofeedback might not improve interoceptive accuracy in the cardiac domain, but effects seem to depend on the specific task applied.