5–200 Hz), and a quality factor of 2 (for 3 dB bandwidths), consistent with those in previous models of human modulation

filtering (Dau et al., 1997), and broadly consistent with animal neurophysiology data (Miller et al., 2002 and Rodríguez et al., 2010). Although auditory neurons often exhibit a degree of tuning to spectral modulation as well (Depireux et al., 2001, Rodríguez et al., 2010 and Schönwiesner and Zatorre, 2009), this is typically less pronounced than their temporal modulation tuning, particularly early in the auditory system (Miller et al., 2002), and we elected not to include it in our model. Because 200 Hz was SNS-032 cost the Nyquist frequency, the highest frequency filter consisted only of the lower half of the half-cosine frequency response. We used a smaller set of modulation filters to compute the C1 and C2 correlations, in part because it was desirable to avoid large numbers of unnecessary statistics, and in part because the C2 correlations necessitated octave-spaced filters (see below). These filters also had frequency responses that were half-cosines on a log-scale, but were more broadly tuned ( Q=2),

with center frequencies in octave steps from 1.5625 to 100 Hz, yielding seven filters. All filtering was performed in the discrete frequency domain, and thus assumed circular boundary conditions. To avoid boundary artifacts, the statistics measured in original recordings were computed as weighted time-averages. The weighting DNA Damage inhibitor window

fell from one to zero (half cycle of a raised cosine) over the 1 s intervals at the beginning and end of the signal (typically a 7 s segment), minimizing artifactual interactions. For 2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase the synthesis process, statistics were imposed with a uniform window, so that they would influence the entire signal. As a result, continuity was imposed between the beginning and end of the signal. This was not obvious from listening to the signal once, but it enabled synthesized signals to be played in a continuous loop without discontinuities. We denote the k  th cochlear subband envelope by sk  (t  ), and the windowing function by w  (t  ), with the constraint that t∑w(t)=1∑tw(t)=1. The nth modulation band of cochlear envelope sk is denoted by bk,n(t), computed via convolution with filter fn. Our texture representation includes the first four normalized moments of the envelope: Mk1=μk=t∑w(t)sk(t),M1k=μk=∑tw(t)sk(t), M2k=σk2μk2=∑tw(t)(sk(t)−μk)2μk2, M3k=∑tw(t)(sk(t)−μk)3σk3,and M4k=∑tw(t)(sk(t)−μk)4σk4k∈[1…32]ineachcase. The variance was normalized by the squared mean, so as to make it dimensionless like the skew and kurtosis. The envelope variance, skew, and kurtosis reflect subband sparsity. Sparsity is often associated with the kurtosis of a subband (Field, 1987), and preliminary versions of our model were also based on this measurement (McDermott et al., 2009).