9The speech-coding breakthrough

FGood hardware, modest results

The multichannel implants of the mid-1980s were a genuine advance over single-channel devices, but most users still could not hold a conversation without lip-reading, and open-set speech scores were highly variable. The hardware seemed adequate — multiple electrodes sitting along the cochlea — so why was performance not better? The answer turned out to lie in how the electrodes were being driven, not in the electrodes themselves.

CThe channel-interaction problem

Early multichannel strategies were largely analogue and simultaneous: several electrodes delivered current at the same time. But the cochlea is a conductive, fluid-filled space, so currents from neighbouring electrodes spread and overlap, summing unpredictably at the nerve. This channel interactionsmeared the very spectral distinctions the multiple channels were meant to provide — the implant had many channels on paper but far fewer effectively independentones. The sceptics' old worry about channel interaction (Module 5) had been real after all; it just needed solving rather than surrendering to.

CThe CIS idea

Blake Wilson and colleagues at the Research Triangle Institute proposed an elegant fix: do not stimulate the electrodes at the same time. Continuous interleaved sampling (CIS) breaks sound into frequency bands, extracts the slowly varying envelope of each, and delivers them as brief biphasic pulses staggered in time, so that only one electrode is ever active at a given instant. Because the pulses never overlap, the currents cannot interact in the same way, and each channel conveys its information cleanly.[1991]

FTThe impact

The result, reported in Nature in 1991, was dramatic and immediate: large improvements in speech-recognition scores for the same recipients with the same implanted hardware, simply by changing the processing strategy. It was the clearest possible demonstration that the bottleneck had been the code, not the wires. Every modern coding strategy — the ACE, SPEAK, HiRes and fine-structure families used by today's manufacturers — descends from or responds to the CIS insight.[1991]

CThe lesson that endures

CIS reframed the whole enterprise. The cochlear implant is not just a set of electrodes; it is an information-delivery problem, and the strategy that converts sound into a pattern of pulses is as important as the electrode array carrying them. This is why a programming audiologist's choices, and the objective measures that inform them, matter so much (the Objective Measures chapter) — and why progress in cochlear implants continues to come substantially from software. Wilson shared the 2013 Lasker Award with Clark and Hochmair largely for this contribution (Module 13).

With the device now able to deliver real speech, the story turns from invention to acceptance: the approvals and the consensus that made implantation standard care — FDA approval and the 1988 NIH consensus (Module 10).