Neural circuits capable of generating multiple outputs are essential for behavioral flexibility, yet their organizational principles remain poorly understood. Using vocal communication in singing mice (Scotinomys teguina), we investigated whether distinct vocal behaviors are controlled by separate pathways or by shared circuits operating under different parametric regimes. We developed a novel behavioral assay (PAIRId---Partial Acoustic Isolation Reveals Identity) that enables precise attribution of vocalizations during social interactions in singing mice. This approach revealed two major vocal modes: loud, temporally patterned songs used for long-distance communication and soft, unstructured ultrasonic vocalizations (USVs) employed during close-range interactions. Despite their dramatic acoustic and contextual differences, both vocal modes share peripheral sound production mechanisms and central neural control by the caudolateral periaqueductal gray (clPAG). We derived a simple mathematical model describing song rhythm as a linear progression of note rates, which captures song motor patterning with just three parameters and accurately predicts song duration across animals and conditions. Using this model, we demonstrate that progressive silencing of clPAG neurons systematically alters specific song parameters before eliminating all vocalizations. Notably, one of these parameters - which controls song termination - also accounts for natural sexual dimorphism in song production. Our findings reveal how differential amplitude and frequency modulation of shared neural circuits produces categorically distinct behavioral outputs and provide a mechanistic basis for how behavioral innovations can emerge through evolutionary tinkering of ancestral neural pathways.