General Description:

Research in the neural encoding laboratory investigates the representation and processing of complex stimuli in the auditory system. One goal is to understand the relationships between the perception of sound and the responses of auditory neurons. Another is to analyze the effects of hearing impairment on the representation and to investigate signal processing for neural prostheses. Some specific examples of our approach:

1. Neural circuits in the brainstem auditory system. How is the brain organized for auditory information processing? How are neurons interconnected and how do they interact?

   Examples of research in this area:

   • What's a cerebellar circuit doing in the auditory system? - a review of the circuitry and functional properties of the dorsal cochlear nucleus, with speculations about the role of this nucleus in the brain.

   • Anesthesia blocks narrowband, but not wideband inhibition in DCN. - anesthesia modifies the response properties of neurons in the dorsal cochlear nucleus (DCN). Unexpectedly, the effect is specific for one kind of inhibitory circuit and not for another, even though both use the same neurotransmitter.

   • Somatosensory input to the DCN carries information abou the position of the pinna - The DCN receives both auditory and somatosensory inputs. The somatosensory ones in the cat respond to stretch of the muscles that are used to move the external ear (pinna). This suggests a role of the DCN in coordinating information about the location of sounds in space.

   • DCN principal cells respond to spectral edges, which requires additional inhibitory effects in DCN - DCN principal cells give a peak discharge rate to rising spectral edges centered near BF. This response cannot be fully explained by the current known circuitry of the DCN.

   
   
2. The representation of complex stimuli in neural responses. How does the activity of neurons in the brain represent the acoustic environment? How do we discriminate between sounds? How can we understand and model the neural representation of sound?

   Examples of research in this area:

   • Discrimination of consonant-vowel syllables can use neural activity at a variety of time scales. - information about the differences between similar speech sounds is contained in a variety of aspects of the spike trains. Here a measure of information sensitive to the differences between stimuli is used to show that stop consonants differ at the formant frequencies, as expected, but also at high frequencies, where the acoustics of the burst at the release of the stop are important.

   • Receptive fields of auditory neurons can be linear or nonlinear. - Auditory spectral receptive fields become nonlinear in some neurons in the cochlear nucleus. A method of constructing receptive fields for spectral shape (i.e. the frequency content of sounds) gives first and second-order receptive fields. These are used to show that neurons in the ventral cochlear nucleus are reasonably linear, i.e. well-represented by first plus second order models, whereas neurons in dorsal cochlear nucleus are frequently nonlinear. This is the first systematic description of the effects of second-order components of receptive fields.

   • Information about sound localization is distributed across neuron types in the inferior colliculus. - Three neuron type can be recognized in the inferior colliculus, based on response maps. These seem to be connected differently to brainstem auditory neurons, suggesting a difference in the representation of different sound localization cues. Analysis of the representations using mutual information shows that some segregation exists, but generally auditory information is distributed broadly across the response types.

3. Studies of stimulus representation in animals with hearing impairment. Acoustic trauma is used to produce a hearing loss resembling a sloping high-frequency hearing loss, typical of older listeners and hearing-aid users.

Examples of research in this area:

• Auditory nerve fibers in impaired ears and recruitment - in impaired ears, sounds become louder more rapidly with intensity than in normal ears. This recruitment phenomenon is a problem for hearing aid design and often limits what can be done. Here, we show that auditory nerve fibers do not behave as expected by the standard model of recruitment. This work suggests that theories of recruitment may have to be revised.

• A model of auditory-nerve responses in ears with hearing impairment - this computational model is intended to be useful in testing hearing-aid signal processing strategies.i

• Discriminability of vowels is enhanced in impaired ears by a conditional amplification strategy - Discrimination of speech sounds based on auditory nerve responses is impaired in ears following acoustic trauma. Here a hearing-aid amplification strategy is suggested which partially restores the normal representation.