When there is no nerve to stimulate
A cochlear implant needs a functioning auditory nerve to carry its signal centrally. Several conditions destroy that path: neurofibromatosis type 2, where bilateral vestibular schwannomas and their removal sever the nerve; congenital cochlear-nerve aplasia; and a cochlea so ossified that no array can be inserted. For these patients, no amount of peripheral cleverness helps — the wire has nowhere to send its message.
Even when the nerve is intact, insertion is not free. Deep or forceful placement causes electrode-array insertion trauma, damaging the basilar membrane and the very spiral-ganglion neurons the device depends on. Hearing-preservation ('soft') surgery and electrode design are an active concern precisely because the neural target is finite and non-renewable.
The auditory brainstem implant
The auditory brainstem implant (ABI) places a small paddle of surface electrodes onto the cochlear nucleus in the lateral recess of the fourth ventricle, one synapse past the missing nerve. It is a remarkable operation — but the results are sobering, and this is where honesty is essential.
For NF2 patients, most ABI users gain awareness of environmental sound and a substantial aid to lipreading; open-set speech understanding without visual cues is the exception, not the rule. Interestingly, non-tumour recipients — children with cochlear-nerve aplasia, or adults whose brainstem was not distorted by tumour surgery — tend to do considerably better, which points the finger at the local anatomy and off-target tissue rather than the concept itself.
Deeper still: the midbrain, and the safety wall
The auditory midbrain implant (AMI) goes one rung higher, penetrating the inferior colliculus, whose central nucleus has a clean, laminar tonotopic map — in principle an ideal place to write frequency. Small first-in-human trials showed that stimulation produces graded pitch percepts and aids lipreading, but open-set speech has remained limited, in part because a single shank samples that laminar map along only one axis, poorly matched to its geometry.
Every central target also runs into a hard physical wall: charge safety. To write a percept you must inject enough charge to recruit neurons, but too much charge — or too high a charge density — damages tissue. The Shannon–McCreery relation marks that boundary and constrains how finely you can steer current at any central site.
Shannon–McCreery safety boundary: charge density per phase D versus charge per phase Q, with k ≈ 1.5–2.0 marking the onset of tissue damage. Fine current steering must live below this line — see the charge-density safety limit and charge-balanced pulses.
The lesson of going central is consistent. Each step up the ladder buys you access when the periphery has failed, but you pay in off-target activation, loss of an easily-reached tonotopic map, and surgical risk. Directly writing to auditory cortex would be the ultimate step — and remains experimental for exactly these reasons.