Activity of neurons in the auditory system is modulated by
auditory context, the slowly changing statistical properties of the
acoustic environment. Standard encoding models, such as the
linear-nonlinear spectro-temporal receptive field (LN STRF), are unable
to account for these relatively slow contextual effects. Several
alternative models for context-dependent coding have been proposed. We
developed a variant of the STRF in which inputs to A1 neurons undergo
nonlinear short-term synaptic plasticity (STP). We compared performance
of LN and STP STRFs using single-unit data collected from A1 of awake
ferrets during presentation of a library of natural sounds. The STP STRF
showed improved explanatory power over the linear STRF, as well as
properties that support a rich encoding of sound information over longer
timescales than traditional models. In ongoing efforts, we are studying
how these properties can change with behavioral state. We are also
comparing the STP STRF to models that incorporate nonlinear gain control
to determine if they explain equivalent or complementary neural response
properties.