Previous research suggests that musical phrase structures may synchronise listeners' autonomic variables. We use a computational method to automatically identify probabilistic phrase arc boundaries from music audio and compare them to physiological envelopes (respiration and RR intervals). Automating the evaluation of synchronisation of autonomic responses to musical phrases enables the empirical evaluation of music's physiology-modulating power. Participants' respiration and RR intervals were recorded while listening to versions of Prokofiev's Gavotte Op.12 No.2. A novel Bayesian dynamic programming algorithm is used to derive phrase boundary credence profiles from the loudness. Increased curve similarity is observed between loudness phrase boundary credences and listeners' physiological signal envelopes, with the degree of response affected by track version. Loudness credence and RR interval entrainment is statistically significant for the original version. We developed a fully automated system for evaluating musical phrase arc-autonomic variable entrainment. Initial findings suggest that phrase structures can affect physiological signals.