Keynote lecture 2
Wednesday, Sep 9, 2015
The representation of surprise in the auditory system: data and models
ELSC, Hebrew University, Jerusalem, Israel
Neurons in the auditory system, starting at least as early as the inferior colliculus (IC), show stimulus-specific adaptation (SSA) - the selective reduction in their responses to a repeated, standard, stimulus that doesn't generalize to other, deviant stimuli. Subcortical SSA is strong in the so-called non-lemniscal pathway, but is weak in the core, lemniscal pathway from the brainstem to auditory cortex. We know that subcortical SSA does not depend on cortical feedback. On the other hand, primary auditory cortex (A1), which shows a substantial amount of SSA, is mostly driven by lemniscal inputs that do not show SSA. This picture suggests that SSA is computed at least twice in the auditory system: once in the IC, and a second time in A1. To test this hypothesis, we studied the responses of neurons in IC, the medial geniculate body (MGB, the auditory thalamus) and A1 in response to a large number of tone sequences and complex sounds. Here I will describe these results and interpret them in the context of adaptation in narrowly tuned modules (ANTM) models, arguably the simplest class of models that show SSA. Such models predict deviant responses that are the same or smaller than the responses to the same sounds in the deviant alone configuration or when presented as part of 'control' sequences. Furthermore, because of the tonotopic organization of the auditory system, such models predict no SSA to appropriately balanced broadband stimuli. I will show that these predictions hold in IC and in MGB, but fail in A1. In particular, the responses to deviants in A1 are larger than expected by the level of adaptation of the afferent inputs, and neurons in A1 show SSA to particularly well-balanced broadband stimuli that fail to evoke SSA in IC and MGB. I will discuss our approaches for unraveling the circuit mechanisms in A1 that may underlie these results.