11th November 2003 - How we hear where sounds are coming from

The November lecture was given by Professor Alan Palmer from the Institute of Hearing Research at the Medical Research Centre, and focused on the neurological processes that allow humans to perceive where sounds are coming from. The presentation started with a summary of the complex task that the ear-brain combination has to undertake in order to make sense of the audible world. Alan compared this to determining the number and location of boats in a lake simply by observing the movement of corks within two channels cut into the bank of the lake. He explained that we make use of a number of cues to divide the incoming sound into separate components including common time onsets, common amplitude variations, common harmonicity, pitch, and spatial position. It was the last of these that was discussed in detail in the presentation.

Alan explained that humans are able to perceive the position of a sound (a process called localisation) mainly through the use of three cues: pinna spectral cues; interaural level differences; and interaural time differences. He showed an overview of the auditory nervous system and pointed out what are thought to be the main areas for detecting each of these cues.

Pinna cues are generally large spectral deviations at relatively high frequencies caused by constructive and destructive interference between the direct sound from the source and reflections from various parts of the pinna (the external ear). He gave an overview of the circuitry that may be used to detect these cues, and showed where in the auditory nervous system this process takes place. However, he admitted that more research is required to accurately determine how the detection is undertaken.

Alan explained that the pinna cues are monaural, i.e. they only depend on a signal reaching one ear. The other cues to location are binaural - based on quantifying differences between the signals that reach each ear. There are a number of advantages to binaural hearing, including improved detection resulting from increased loudness, simpler removal of interference from echoes, improved detection of sound in a background noise, improved spatial localisation, and improved detection of auditory motion.

The two main binaural cues are interaural time difference (ITD) and interaural level difference (ILD). Alan showed that these cues are created due to the fact that we have two ears, one on each side of our head. The physical distance between the ears causes sounds to the side to reach one ear before the other (giving the ITD), and the shadowing of the head causes sounds to the side of the head to be attenuated at the further ear (giving the ILD).

Alan explained that the ILD seems to be detected by a section of the brain that examines the differences between the signals reaching the ears. He showed that the neural information is passed from the ear via a relay to a section in the centre of the brain. Here, a signal from one ear acts as a positive signal (an excitation) and a signal from the other ear acts as a negative signal (an inhibition). This means that if the signals are the same level, they cancel and nothing is passed on. On the other hand, if the positive signal is a higher level than the negative signal, a signal corresponding to the magnitude of this difference is passed on.

The remainder of the presentation focused on the detection of the ITD, the area of research that Alan and his team is involved in. He first commented on the sensitivity of ITD perception. Amazingly, humans can detect ITDs as small as 20µs, and with training and experience this can even be as low as 4µs (1/250000th of a second). He explained that the basis for the auditory time discrimination is the fact that the neural spikes from the ear lock to a specific phase of the incoming signal, and therefore fire at approximately the same point on every cycle. These phase-locked neural spikes are then passed along the auditory nervous system to be processed in the centre of the brain.

Alan revealed that until recently it was hypothesised that the ITD was detected using a pair of delay lines with coincidence detectors, a model suggested by Lloyd Jeffress in the late 1940s. An example of this is shown in the figure below. This process is a simple method of converting the ITD to a form of spatial map, where a signal arriving simultaneously at the ears will cause a high response in the middle of the delay line, and a signal arriving earlier at either ear will cause a high response correspondingly towards the appropriate end of the delay line.

A simplified diagram of the Jeffress delay line and coincidence model

It appears from neurophysiological research that the Jeffress model may be an accurate representation of the ITD detection process in barn owls and chickens. However, for most mammals, it appears that the auditory neural system is a little different, and there is not good evidence for a delay line. In fact, Alan showed that recent research has suggested that the detection of ITDs involves a process of inhibition, rather than the delay lines required for the Jeffress model.

A question that Alan has been working on with his colleagues is whether there are a range of detectors tuned to respond maximally to different locations in space. They expected that if this was the case, the detectors would be tuned to ITDs across a plausible range of ±700µs (based on the maximum ITD possible due to the distance between the ears), and that the majority of these detectors would be tuned to respond maximally to signals around an ITD of 0 (as this is where our ITD perception is known to be most sensitive). However, they found that neither of these predictions was true.

In fact, it appeared that there are a range of detectors that respond best to different frequencies, and that they respond best to an ITD that is dependent on the best frequency. For example, for low frequency stimuli they respond maximally to an ITD that is away from zero to one side, and for higher frequencies they respond maximally to an ITD that is closer to zero. By plotting the results of a number of these detectors on a common graph, it was found that the frequency dependent tuning resulted in the slope of the plotted curve being steepest close to an ITD of 0 for all of the detectors. This meant that a small change of ITD around 0 resulted in a maximum change in the response, as opposed to the maximum output from the detector being at an ITD of 0.

Alan showed that this process occurs in parallel for the two ears, and that taking the difference between the results for these two systems gives a result that is very sensitive to variations in ITD around a value of 0, and whose sensitivity to changes decreases as the ITD increases in either direction. Assuming a certain amount of inaccuracy and variability due to the biological processes involved, it was demonstrated that the sensitivity predicted by this hearing model matched psychoacoustical results fairly well. The lecture was followed by an interesting discussion about how the knowledge gained from neurological studies of auditory localisation relates to the experience of listening to reproduced sound. Finally, the audience thanked Alan for his fascinating presentation with warm applause.

Russell Mason