Structure of the model. \(\bf(A)\) The organization of the model AC mimics that of the primate AC. Likewise, it is subdivided into 13 fields—three core, eight belt, and two parabelt fields \citep{Baumann2013,Kaas2000,Hackett2014}. The model also includes two subcortical fields representing inferior colliculus (IC) and thalamus. Each field contains 16 columns comprising interacting excitatory and inhibitory mean-field populations. Interactions between fields are indicated by black and blue arrows which signify high and low density of connections, respectively. The cortical connections are bi-directional and result in multiple streams of feedforward and feedback activation. The input to AC is guided along direct connections from thalamus to each of the three fields of the core area. \(\bf(B)\) The connections between different fields and the corresponding synaptic weights between the columns are transformed into weight matrices; here, the matrix \(W_{\tiny\mbox{ee,lat}}\) is shown as an example. The values of the matrix elements indicated by the colour bar to the right represent the strength of the connections between any two fields: positive values stand for excitatory, negative values for inhibitory connections. The intra-cortical elements are computed based on a Gaussian distribution according to \(Q(x)=r\exp[-(x+s\ \mathcal{N}(0,1))^{2}/\sigma]\) with amplitude \(r\) and width \(\sigma\), and where x runs orthogonally to the diagonal of any field. The parameter \(s\) determines the level of stochasticity of the connections; \(\mathcal{N}(0,1)\) is the standard normal distribution. Note that the connections between IC and thalamus as well as between thalamus and the three core fields are purely tonotopic.