It has long been hypothesized that principal neurons in the dorsal cochlear nucleus (DCN), physiological type IV neurons, are wideband spectral feature detectors. The above hypothesis is based on previous systems analysis and other approaches, which show that type IV neurons are nonlinear. Attempts to characterize spectral processing by type IV neurons with 1 ^st- and 2^nd- order approaches have failed. One possible role of the DCN feature selectivity is encoding of features in head related transfer functions (HRTFs), spectral cues for sound localization created by the pinnae. A previous hypothesis-guided approach using a specific class of stimuli, notched noise and band pass noise at varying frequencies, showed that type IV neurons respond with a peak in firing rate to rising spectral edges at their best frequency (BF). This suggested that the DCN type IV neurons could encode sound location information through its excitatory responses to the rising edges present in HRTFs. However, it is not clear if there might be other features that DCN type IV neurons might be encoding. Further, most studies of the DCN, like the ones mentioned above, have looked exclusively at the spectral processing aspects with temporally static stimuli unlike natural conditions.

This study attempts to characterize spectral and temporal processing by DCN principal neurons using systems analysis approaches. A nonlinear spectral receptive field model (a level dependent weight model; LDWM) is derived for DCN principal neurons based on rate responses to random spectral shape (RSS) stimuli of various contrasts and sound levels. These models predict responses to large contrast (natural sound like) stimuli better than earlier used 2 ^nd-order models and hence are better descriptors of the true receptive fields of the true fields of type IV neurons. However, they are poorer than 2^nd-order models for small contrast stimuli as they lack explicit frequency interactions. These models also predict different degrees of edge sensitivity of type IV neurons as shown previously with notched noise stimuli.

Second, to study spectral feature selectivity of DCN type IV neurons from a general perspective online rate maximization experiments were done using RSS-like stimuli. A gradient ascent algorithm was employed. Results from those experiments show tuning of DCN type IV neurons generally to rising spectral edges in a narrow frequency region round the BF. Thus starting from a general set of broadband sounds, these results show that DCN type IV neurons are selective to rising spectral edge shapes in wideband sounds, based on their rate responses.

Finally temporal envelope processing by DCN principal neurons is characterized using random temporal shape (RTS) stimuli, with wideband and narrowband carriers. First and 2^nd-order models of temporal processing show that DCN principal neurons are fairly linear for small fluctuation size stimuli. The nonlinear temporal processing behavior of DCN type IV neurons is based on the fact that they show nonmonotonic envelope synchronization. The nonmonotonicity in envelope synchronization can be explained by 2^nd-order interactions.

A prediction about the function of the DCN is made based on combining their spectral and temporal receptive fields assuming separability. DCN principal neurons are predicted to be tuned to fast (order of 1/16^th octaves/ms) movements of edges (or notches) around their BFs. Such edge or notch movements occur due to pinna or head movements that cause notches in HRTFs to move in frequency. This tuning to edge movements predicts that the DCN might be encoding pinna or head movements.