For each neuron the reaction to sine tones from 500Hz to 32kHz was tested at sound levels from 30 to 90 dB spl. Type of response, best-frequency and latency time were recorded and compared within the pairs. In spite of the tonotopic organisation of the auditory cortex, best frequencies of neighbouring neurons differed by about1/2 octave. This shows that the projection from the cochlea to the tonotopic fields in the cortex is not very precise. Furthermore,the difference in latencies of the responses ranged from zero to about 80 ms with a mean difference of 15 ms. This suggests a complex pre-computation in lower brain regions before sensory information is fed into the cortical layers.

Networks which are structured in a Hebbian learning manner are expected to show the property that synaptic links between cells with similar responses are usually stronger than those between cells which seldom get excited simultaneously, since synaptic weights should increase on coincidental activation of the pre-and postsynaptic neuron. We tested this hypothesis by comparing the difference of best-frequencies in unconnected and interconnected cell pairs. Best-frequencies of unconnected neurons differed by about 0,7 octaves. Pairs which shared a common input were more similar: Their best-frequencies had a mean difference of 0,5 octaves and excitatorily connected cells differed only by about 0,4 octaves. Also response-latency differences within directly connected pairs were much smaller than in other pairs (EXC = 7 ms, CI = 16 ms, NC = 12 ms). These results are consistent with the hypothesis that the cortical network is structured by some learning process.

Figure: Best-frequencies and response-latencies of neighbouring neurons are plotted against one another, showing that connected cells are more similar in their properties than unconnected pairs (NC).

Published in: Gene - Brain - Behaviour, Thieme 1993