, 2011). Synaptic depression also shows temporal asymmetry similar to that observed here (Hosoya et al., 2005, Dobrunz et al., 1997 and Chung et al., 2002). Gain control is primarily useful for adapting the limited dynamic range of a neuron to the statistics of the stimulus. When spectrotemporal contrast is low, firing rates are sensitive to smaller changes within their spectral “region of interest” than under higher-contrast conditions. Thus, the representation of stimulus space is effectively expanded under low-contrast stimulation and compressed under high-contrast stimulation. Decitabine cost Consequently, gain control should improve the ability of individuals

to detect small changes in low-contrast sounds. Indeed, a related phenomenon has been demonstrated in the adaptation to reverberation, whereby listeners are better able to discriminate

(low-contrast) reverberant words when embedded within a reverberant context sentence than within a (high-contrast) anechoic context (Watkins, 2005), an effect that is also frequency-band specific (Watkins and Makin, 2007). Perceptual adaptation is not, however, complete, as a general increase in the spectrotemporal contrast of speech leads to selleck screening library demonstrable gains in intelligibility (Steeneken and Houtgast, 1980, van Veen and Houtgast, 1985 and Miller et al., 1999). Our data predict that perceptual adaptation to stimulus contrast should be observable with nonspeech stimuli as well. Neurons in the visual system are subject to contrast gain control, which is thought to be desirable for efficient coding of natural images (Schwartz and Simoncelli, 2001). Since the contrast of natural images is correlated across space and time, normalization by stimulus contrast reduces the redundancy of aminophylline the neural code (Barlow, 1961 and Vinje and Gallant, 2002). The contrast of a complex visual stimulus can be defined as σI/μI, which is strongly related to the two contrast measures that we have shown to determine auditory gain control (σL, Figure 5A; σP/μP, Figure S4C).

Auditory gain control may therefore have a similar redundancy-reducing effect. Although the ensemble (i.e., long time scale) distributions of natural sounds have been explored ( Attias and Schreiner, 1997, Escabí et al., 2003 and Singh and Theunissen, 2003), a deeper understanding of the relationship between contrast gain control and the statistics of natural sounds will require a characterization of natural sound level distributions at the temporal scales over which gain control operates. We show that when stimulus level statistics are not uniform across the spectrum, gain control is also unevenly applied to neurons, depending on their frequency tuning. A spectrally limited band of high contrast has the greatest compressive effect on neurons if their tuning curves overlap this band.