In the first scenario, the efficacy to elicit a response pattern would be independent of the presence of the second component, whereas in the second scenario the switch would

be defined by the specific ratio of the two components making up the mixture. To test this hypothesis, we again identified local populations generating two response modes (n = 5, in 2 mice) and selected two basis sounds exciting each of the modes. We synthesized mixtures of the two basis sounds with seven equally spaced mixture ratios and the individual mixture components in isolation (i.e., one of the basis sounds faded to silence). We compared the patterns of response to the isolated components and to the mixtures and computed the corresponding clustered similarity matrices (Figures 5H and S5). The sound level at which the transition

occurred depended on the specific sound and the local population. Importantly, Everolimus in all cases we found sound levels where both components of the mixture in isolation elicited a reliable response. However, when both components are presented at the same time one of the two response modes appeared to be dominant and only one of the patterns was excited (Figure 5H). Hence, instead of an additive response, the local network falls in a highly Selleckchem mTOR inhibitor nonlinear manner in either one of the two response modes. This indicates that the choice of one mode or the other is a winner-take-all decision, which may result from competitive interactions between neuronal populations. Our observations show that local populations of the auditory cortex are constrained to few response modes which encode a small number of sound categories enclosing several sounds. This implies that local populations are highly limited in their capacity to discriminate a large number

of sounds. Yet at the level of the organism, sound discrimination does not show such constraints. How this apparent paradox could be resolved became evident when we probed various local populations within and across mice in several primary auditory fields (Figure 6A). In each case, the different local populations categorized different sets of sounds, suggesting that different local populations provide complementary information the to unambiguously encode a large number of sounds. To quantitatively assess this observation, we plotted response similarity matrices for a selection of 15 clearly distinct sounds (excluding mixtures and different sound levels; Figure 6B), in which the sound order is fixed (i.e., no clustering was performed; Figure 6C, top). In these plots, the sounds giving rise to a reliable response or being grouped in different modes differ from one population to the next. Thus, different populations are discriminating different sets of sounds.