1. The contribution of the midbrain periaqueductal gray (PAG) to the central regulation of vocalization was investigated by analyzing the electromyographic (EMG) changes in respiratory, laryngeal, and oral muscles evoked by microinjection of D,L-homocysteic acid (DLH) in the PAG of unanesthetized, precollicular decerebrate cats. Moderate to large (6-40 nmol) doses of DLH evoked natural-sounding vocalization as well as increases in inspiratory depth and respiratory rate. 2. Two basic types of vocalization were evoked, each associated with a distinct and characteristic pattern of respiratory, laryngeal and oral EMG changes. Type A vocalization (voiced sounds such as howl/mew/growl) was characterized by excitation of the cricothyroid (CT) and thyro-arytenoid (TA) muscles, and inhibition of the posterior crico-arytenoid (PCA) muscle, whereas type B vocalization (unvoiced hiss sounds) was characterized by excitation of the PCA and TA muscles and no significant activation of the CT muscle. In addition, stronger expiratory (external oblique, internal oblique, internal intercostal) EMG increases were associated with type A responses, and larger increases in genioglossus and digastric muscle activity were associated with type B responses. 3. Microinjections of small doses of DLH (300 pmol-3 nmol), also evoked patterned changes in muscle activity (usually without audible vocalization) that, although of lower amplitude, were identical to those evoked by injections of moderate to large DLH doses. In no such experiments (175 sites) were individual muscles activated by small dose injections of DLH into the PAG. Further, type A vocalization/muscle patterns were evoked from PAG sites caudal to those at which type B vocalization/muscle patterns were evoked. 4. Considered together these results indicate: that the PAG contains topographically separable groups of neurons that coordinate laryngeal, respiratory, and oral muscle patterns characteristic of two fundamental types of vocalization and that the underlying PAG organization takes the form of a representation of muscle patterns, rather than individual muscles. 5. The patterns of EMG activity evoked by excitation of PAG neurons were strikingly similar to previously reported patterns of EMG activity characteristic of major phonatory categories in higher species, including humans (e.g., vowel phonation, voiceless consonant phonation). These findings raise the possibility that the sound production circuitry of the PAG could well be utilized by cortical and subcortical "language structures" to coordinate basic respiratory and laryngeal motor patterns that are necessary for speech.