Besides phonetic information, speech contains indexical cues that are informative for, amongst other things, identifying who is talking. Vocal-tract length (VTL), which is related to the size of the speaker, is one of the important cues associated with speaker identity. Previous studies have shown that VTL perception is strongly impaired in most cochlear-implant (CI) users. Here we are exploring potential causes for this observation.

When speech is transmitted through a CI, it is first transformed into a series of electric pulses through a processing strategy. These processing strategies have been optimised for the transmission of phonetic information through the constraints of electrical stimulation. However, the strategies might not be optimal for other types of information such as VTL. Furthermore, the pulses generated by the strategy are then used to stimulate the auditory nerve by electrical stimulation through a rather conductive medium. This results in a large spread of excitation where the activated neurons are not necessarily located close to the stimulating electrode, thereby reducing the effective spectral resolution available to the listener. Consequently, both the processing strategy and the current spread at the electrode-neuron interface could play a role in reducing the information available to the listener and it can be difficult to disentangle the two.

To disentangle the contributions of these two mechanisms, we recorded the electrical outputs produced by a Cochlear Nexus 7 processor using the MP3000 coding strategy. These electrical outputs were then resonified using a vocoder that allowed us to control the amount of (simulated) current spread by adjusting the spectral width of the carrier. By comparing the VTL discrimination performance obtained in the original stimuli, to those obtained through vocoding with and without current spread, we could estimate how much the coding strategy alone and the electrode-neurone interface each contribute to the loss of sensitivity.

We collected VTL sensitivity from 16 normal-hearing adults across two variations of the experiment. Our results indicate that the MP3000 strategy preserves VTL information remarkably well, but that current spread quickly reduces the performance.

These results now need to be confirmed in actual CI users, where other measures of spectral resolution and current spread can be obtained using tests such as spectro-temporal ripple discrimination, as well as more objective measures of current flow based on eCAP measures.