OWL AUDITORY SYSTEM

The cochleas transduce sound into neuronal firing patterns distributed across the fibers of the two auditory nerves. (a) Describe the primary stages by which interaural time difference (ITD) is calculated using these two sets of signals and how the phase ambiguity problem is solved (be sure to describe why this problem occurs, and what "characteristic delay" means). (b) What is one similarity between the neural solution to the phase ambiguity problem for interaural time delay and the neural solution to the aperture problem for pattern motion?

The barn owl (Tyto alba) uses interaural time differences (ITDs) for localization of sound in the horizontal plane and interaural intensity differences (IIDs) for localization in the vertical plane (Konishi et al., 1988). The pathways for time disparity and intensity disparity are separate: binaural nuclei in the brain stem are already segregated into two groups: those that are sensitive to time disparity, and those sensitive to intensity disparity (Konishi et al., 1988). The time pathway starts with the cochlear nucleus magnocellularis (NM), whereas the intensity pathway starts with the cochlear nucleus angularis (NA). Neurons in the NM have phase-locked responses, i.e. they tend to fire at a particular phase angle of a sound. When a neuron is selective for a particular time difference/disparity between the signals coming from the two ears independent of stimulus frequency, the neuron is said to have a characteristic delay (Konishi et al., 1988).

The first site of binaural convergence is the nucleus laminaris (NL), which receives input from the left and right NM, and is also selective for interaural time differences. NM and NL are the first and second stations of the time delay pathway, respectively.

a) How Interaural Time Delay (ITD) is calculated:

As stated above, NM neurons have phase-locked responses, therefore helping identify the timing of a stimulus, and the NL receives bilateral input from the (left and right) NM. In addition, evidence suggests that NL neurons act as coincidence detectors (see Figure 7A) (Konishi et al., 1988). Assuming the sound came from the left side of space, the left NM will be activated before the right NM (= ITD). To encode the ITD, the nucleus laminaris would need to match two signals, one of which however arrives delayed at one NM. Since NL neurons fire most strongly in response to the coincidence of a left and right spike, “delay lines” are necessary to facilitate the neuron’s response to both the left and right NM inputs: the axon length to binaural neurons A, B, C and D increases asymmetrically on the left input path versus the right input path – i.e. the axon path increases systematically for left inputs to units D to A, whereas for inputs from the right, it increases systematically for units A-D (decreasing for units D-A). When binaural disparities (sounds coming from either left or right first) exactly compensate for this asymmetry, the neuron fires maximally. This allows NL neurons to respond selectively to specific ITDs. The place of the neuron thus codes for ITD. The anatomic characteristics of the Nucleus Laminaris support this model.

Figure 7:

A) Coincidence detector model:

The phase ambiguity problem arises from the fact that the cochlea filters sound into different frequencies and that the left and right nuclei magnocellulari are sharply frequency tuned, meaning that the noise signal is broken down into multiple frequency components. This means that the NL must derive a time disparity from phase disparities in separate spectral components of the noise. Because of this, the NL doesn’t know how to match the peaks of the binaural spikes in order to find out which spike (from which NM) occurred first. The NL responds well to several ITDs, both negative and positive, indicating that it cannot tell the difference between “left ear first” and “right ear first”. Also, a given time disparity elicits the same rate and timing pattern of discharge in neurons tuned to different time disparities. Hence NL is incapable of signaling the exact time delay, it only signals that there is a time delay between the signals coming from the two ears. This is obviously problematic, since if the left NM signal arrives at t = 0 μs while the right NM signal arrives at t = +20 μs, the NL or some structure should ideally signal that the sound came from the left side of space.

A solution to the problem is achieved in the ICc lateral (inferior colliculus central nucleus, lateral part), the final stage of processing involved in calculating the ITD: integrating across frequencies (just like in the aperture problem for motion, integration of information across receptive fields is required). The NL projects topographically to the ICcl. The ICc lateral contains neurons selective for different frequencies and time disparities. Neurons with the same characteristic delay preference are organized in columns. Therefore, if multiple neurons sensitive to different frequencies fire in response to the same ITD, (e.g. a column of activation in the ICc lateral), the neurons that fire the most will indicate the true time delay / ITD . The column with maximum activation therefore will be selected to determine the interaural time delay. Columns with weaker activation will be ignored, since they do not signal the true time delay.

b) The nucleus laminaris combines spikes from the left NM and right NM to individual frequencies (i.e. for individual spectral components separately) - since the cochlea divides the incoming sound into its component frequencies. The NL is unable to determine the true ITD for the entire noise signal, because it only compares spikes to individual frequencies. This resembles the aperture problem for pattern translation, which is similarly caused by the availability of only limited bits of information (i.e. the NL lacks “the whole picture”, just like V1 neurons lack the whole picture due to their small receptive fields). Just like the aperture problem is solved by combining signals across space/ receptive fields, the phase ambiguity problem is solved by combining signals across the entire frequency spectrum.