1. Azimuth and sound pressure level (SPL) tuning to noise stimulation was characterized in single-unit samples obtained from primary auditory cortex (AI) and in areas of the medial geniculate body (MGB) that project to AI. The primary aim of the study was to test the hypothesis that AI is an important site of synthesis of single-unit responses that exhibit both azimuth sensitivity (tendency for directionally restricted responsiveness) and nonmonotonic (NM) level tuning (tendency for decreased responsiveness with increasing SPL). This was accomplished by comparing the proportions of such responses in AI and MGB. 2. Samples consisted of high-best-frequency (BF) single units located in MGB (n = 217) and AI (n = 216) of barbiturate-anesthetized cats. The MGB sample was obtained mainly from recording sites located in two nuclei that project to AI, the ventral nucleus (VN, n = 118) and the lateral part of the posterior group of thalamic nuclei (Po, n = 84). In addition, a few MGB units were obtained from the medial division (n = 8) or uncertain locations (n = 7). Each unit's responses were studied using noise bursts presented from azimuthal sound directions distributed throughout 180 degrees of the frontal hemifield at 0 degrees elevation. SPL was varied over an 80-dB range in steps of < or = 20 dB at each location. Similarities and differences in azimuth and level tuning were evaluated statistically by comparing the AI sample with the entire MGB sample. If they were found to differ, the AI, VN, and Po samples were compared. 3. Azimuth function modulation was used as a measure of azimuth sensitivity, and its mean was greater in AI than in MGB. NM strength was defined as the percentage reduction in level function value at 75 dB SPL and its mean was greater in AI (showing a greater tendency for decreased responsiveness) than in MGB. Azimuth-sensitive (AS) NM units were identified by jointly categorizing each sample according to both azimuth sensitivity (sensitive and insensitive categories) and NM strength (NM and monotonic categories). AS NM units were much more common in the AI sample than in any of the MGB samples, suggesting that some such responses are synthesized in AI. 4. A vast majority of AI NM units have been reported to be AS, showing a preferential association (linkage) between these two response properties. This finding was confirmed in AI, but was not found to be the case in MGB. This suggests that a linkage between these response properties emerges in the cortex, presumably as a result of synthesis of NM AS responses. Although the functional significance of the linkage is unknown, NM responses may reflect excitatory/inhibitory antagonism that provides AS AI neurons with sensitivity to stimulus features beyond that which is present in MGB. 5. Breadth of azimuth tuning of AS cells was measured as the portion of the frontal hemifield over which azimuth function values were > 75% of maximum (preferred azimuth range, PAR). PARs were broadly distributed in each structure, and mean PAR was narrower in AI than in MGB. A preferred level range (PLR) was defined for NM level functions as the range over which values were > 75% of maximum, and mean PLRs were similar in each sample. There was a weak, but significant, positive correlation between PARs and PLRs in AI but not in MGB. This further suggests a linkage between azimuth and level tuning in AI that does not exist in MGB. 6. Best azimuth (midpoint of the PAR) was used to classify cells as contralateral preferring, ipsilateral preferring, midline preferring, or multipeaked. Samples from AI and MGB exhibited similar distributions of these categories. Contralateral-preferring cells represented a majority of each sample, whereass midline-preferring, ipsilateral-preferring, and multipeaked cells each represented smaller proportions. This suggests that the azimuth preference distribution in AI largely reflects that in MGB. 7. A best SPL was defined as the midpoint of the PLR. This wa