The gesture-speech physics theory suggests that there are biomechanical interactions of the voice with the whole body, driving speech to align fluctuations in loudness and F0 with upper-limb movement. This exploratory study offers a possible falsification of the gesture-speech physics theory, which would predict effects of upper-limb movement on voice as well as respiration. We therefore investigate co-movement expiration. Seventeen participants were asked to produce a continuous exhalation for several seconds. After 3s, they execute one of five within-subject movement conditions with their arm with and without a wrist weight (no movement, elbow flexion, elbow extension, internal arm rotation, external arm rotation). We analyzed the smoothed amplitude envelope of the acoustic signal in relation to arm movement. Compared to no movement, all four movements lead to higher positive peaks in the amplitude peaks, while weight did not influence the amplitude. We also found that across movement conditions, positive amplitude peaks are structurally timed relative to peaks in kine-matics (speed, acceleration). We conclude that the reason why upper-limb movements affect voice loudness is still best understood through gesture-speech physics theory, where upper-limb movements affect the voice directly by modulating sub-glottal pressures. Multimodal prosody is therefore partly literally embodied.