Human interpersonal coordination can yield synchronization at multiple timescales, including behavioral (auditory-motor) and physiological (respiratory and cardiac) levels; yet the causal relationship among these levels is poorly understood. By comparing dyadic melody perception and production, we demonstrate that physiological synchrony is not merely a byproduct of shared perception, as it increases significantly during joint production relative to joint perception or to silence. Perturbing dyads' behavior or respiration revealed distinct causal effects: respiratory perturbations impaired both dyadic respiratory and behavioral synchrony, whereas auditory-motor perturbations disrupted only dyadic behavioral synchrony. Individual differences further shaped synchrony: partners who shared similar spontaneous rates achieved better behavioral synchrony, and partners with more similar resting heart rates exhibited stronger cardiac synchrony in joint production. These findings disentangle the relationships among levels of human synchrony, reveal directional entrainment processes between respiratory and behavioral synchrony, and highlight the pivotal role of individual differences in interpersonal coordination.