Two fundamental aspects of frequency analysis shape the functional organization of primary auditory cortex. For one, the decomposition of complex sounds into different frequency components is reflected in the tonotopic organization of auditory cortical fields. Second, recent findings suggest that this decomposition is carried out in parallel for a wide range of frequency resolutions by neurons with frequency receptive fields of different sizes (bandwidths). A systematic representation of the range of frequency resolution and, equivalently, spectral integration shapes the functional organization of the iso-frequency domain. Distinct subregions, or "modules," along the iso-frequency domain can be demonstrated with various measures of spectral integration, including pure-tone tuning curves, noise masking, and electrical cochlear stimulation. This modularity in the representation of spectral integration is expressed by intrinsic cortical connections. This organization has implications for our understanding of psychophysical spectral integration measures such as the critical band and general cortical coding strategies.

1. The physiology and topography of single neuron responses along the isofrequency domain of the middle- and high-frequency portions [characteristic frequencies (CFs) greater than 4 kHz] of the primary auditory cortex (AI) were investigated in the barbiturate-anesthetized cat. Single neurons were recorded at several locations along the extent of isofrequency contours, defined from initial multiple-unit mapping. For each neuron a high-resolution excitatory tuning curve was determined, and for some neurons high-resolution two-tone tuning curves were recorded to measure inhibitory/suppressive areas. 2. A physiologically distinct population of neurons was found in the dorsal part of cat AI. These neurons exhibited two or three distinct excitatory frequency ranges, whereas most neurons in AI responded with excitation to a single narrow frequency range. These were called multipeaked neurons because of the shape of their tuning curves. At frequencies between the excitatory regions, the multipeaked neurons were inhibited or unresponsive. 3. Multipeaked neurons exhibited several distinct threshold minima in their frequency tuning curves. Most of the multipeaked neurons (88%) displayed two frequency minima, whereas the rest exhibited three minima. 4. The frequency separation between threshold minima was less than 1 octave in 71% of the double-peaked neurons recorded. Occasionally, the frequency peaks of these neurons closely corresponded to a response to second and third harmonics without a response to the fundamental frequency. 5. Multipeaked neurons exhibited a wide range of total bandwidths (highest excitatory frequency minus lowest excitatory frequency expressed in octaves). Bandwidths of the isolated peaks within the same neuron were also quite variable. 6. Response latencies to tones with frequencies within each peak of a multipeaked neuron could vary considerably. In 71% (17) of the neurons, tones corresponding to the high-frequency peak (CFh) elicited a longer response latency (greater than 4 ms) than those corresponding to the low-frequency peak (CF1). 7. Inhibitory/suppressive bands, as demonstrated with a two-tone paradigm, were often present between the peaks. Typically, neurons with excitatory peaks of similar response latencies showed an inhibitory band located between the peaks. 8. Ninety percent of the topographically localized multipeaked neurons were in the dorsal part of AI (greater than 1 mm dorsal to the maximum in the sharpness-of-tuning map). Although these neurons were restricted to dorsal AI, only 35% of neurons in this region were multipeaked. 9. Multipeaked neurons could show decreased response latencies and thresholds to two-tone combinations.(ABSTRACT TRUNCATED AT 400 WORDS)