Cochlear Implant Atlas
CI Atlas · Hearing Music Through an Implant · Module 04

4The Hardest Part: Pitch and Melody

Of all of music's dimensions, pitch is the one the implant degrades most. Place-pitch is coarse and distorted, temporal-pitch saturates within a few hundred hertz, and the casualty is melody: strip away rhythm and many recipients recognise familiar tunes barely above chance.

FTwo ways to hear pitch, both broken

The healthy cochlea encodes pitch in two complementary ways: PLACE (which point along the tonotopic basilar membrane vibrates) and TIMING (the rate at which auditory-nerve fibres phase-lock to the waveform). Music exploits both. An implant inherits a degraded version of each. Place-pitch is delivered by a handful of electrodes (typically 12-22) spanning a cochlea that natively resolves thousands of frequency channels, so the place axis is grossly undersampled. Temporal-pitch is delivered by the stimulation rate or the rate of envelope modulation, but the percept stops rising with rate above roughly 300 Hz - the electrical pitch saturates well below the range music lives in. Because both channels are compromised, pitch is the dimension where the gap between implant hearing and normal hearing is widest.[2014][2004]

How the implant degrades both pitch channels

PLACE channels>3000native12–22implantRATE pitch (Hz)~4 knative~300implantheights log-scaled

Pitch reaches the brain through two routes, and the implant narrows both. On the place axis the healthy cochlea behaves like more than 3000 finely tuned frequency channels arrayed along its length; the implant substitutes just 12–22 electrodes, collapsing that resolution by more than two orders of magnitude. On the rate axis the normal auditory nerve can lock to fine timing up to several kilohertz, but electric rate pitch saturates near 300 Hz, so faster pulse trains stop sounding higher. With both channels capped, the residual pitch the listener receives is coarse on place and shallow on rate. Schematic.

TWhy place-pitch is ambiguous

Tonotopy is only loosely preserved after implantation: the array sits in the scala tympani, often does not reach the apical low-frequency region, and the map between an electrode and the cochlear place it stimulates is distorted and shifted relative to a normal ear. Current spread is the deeper problem - each electrode injects current that excites a broad swath of neurons, so neighbouring electrodes overlap heavily and the effective number of independent pitch channels is far smaller than the electrode count. The result is that two notes a semitone apart may stimulate overlapping neural populations and sound the same. Pitch-ranking studies show many CI users perform near chance for 1-semitone differences, the very resolution Western melody depends on. Reusing low-frequency apical sites is limited by both electrode insertion depth and surviving spiral-ganglion neurons, so the place axis cannot simply be 'turned up' in resolution.[2014][2008]

Frequency JND: cochlear implant vs normal hearing

08152330JND (% of frequency)Normal (typical)Normal (worst)CI (best)CI (typical)CI (worst)
ListenerCI (worst)Frequency JND30%

The frequency JND is the smallest pitch change a listener can reliably detect. A normal-hearing ear is exquisitely sharp, resolving changes well under 1% — about 0.2 to 0.5% of the reference frequency. A cochlear implant is far coarser: the typical electric JND is around 10%, with the range running from roughly 1% in the best users to about 30% in the poorest. A 10% step is more than an entire musical tone, which is why precise melody and intonation slip through the implant even when speech does not. Illustrative.

TWhy temporal-pitch saturates

When place cues are weak, the implant must lean on temporal cues - the periodicity of the envelope or the stimulation rate - to convey pitch. Electrically evoked pitch does rise with rate at low rates, but the percept plateaus near 300 Hz; faster rates no longer sound higher. Normal pitch tracking extends to several kHz. This ceiling sits right inside the musical range: middle C is about 262 Hz and the notes immediately above it cross the saturation point, so much of a melody's pitch movement falls where electrical pitch can no longer change. Just-noticeable frequency differences for CI users average roughly 10% (and range from ~1% to 30%), an order of magnitude worse than the well-under-1% of normal hearing - far too coarse for reliable melody.[2014][2004]

Melody recognition: with rhythm vs rhythm removed

0255075100correct (%)CI usersNormal hearing
ListenerNormal hearingWith rhythm cues98%Rhythm removed92%

Ask a cochlear-implant listener to name a familiar tune and they often succeed — but largely by reading its rhythm. With the rhythm cue intact, recognition sits well above chance (~60% here). Strip the rhythm out so that only the pitch contour distinguishes one melody from another, and CI performance falls toward chance (~12%), exposing how little pitch information the implant carries. Normal-hearing listeners barely change, staying near ceiling with or without rhythm because their pitch coding does the work. Tap a bar group, or a legend item to isolate a condition. Illustrative.

CThe consequence: melody collapses, contour barely survives

Melodic contour identification - simply judging whether a five-note pattern rises, falls or stays flat - is highly variable across CI users (14-91% correct in one classic study) and improves only as the note intervals are widened. Familiar-melody recognition with rhythm and lyrics removed is the harshest test: normal-hearing adults score far higher, and many implant users fall near chance, identifying tunes only by guessing. Performance jumps when rhythm cues are restored, because recipients identify the song by its rhythmic signature rather than its pitch contour - a compensation, not a recovery of melody. Clinically: counsel recipients that pitch and melody are the slowest, least complete part of music to return, and that contour-identification training can yield measurable, generalising gains.[2007][2002][2005][2004]

Case 29.4 · The Hardest Part
A 54-year-old post-lingually deafened CI user, six months after activation, reports she can 'feel' songs and clap along perfectly, and recognises 'Happy Birthday' instantly - yet when her therapist plays familiar tunes on a keyboard with every note the same length, she cannot name a single one and her guesses are random.

What best explains the dissociation between her preserved song recognition and her failure on the equal-duration melody test?

Self-assessment — Module 43 questions
Question 1

Why does electrically evoked temporal pitch fail to convey most musical melodies?

Question 2

What is the main reason place-pitch is ambiguous in cochlear implant users?

Question 3

Compared with normal hearing, the average just-noticeable frequency difference for CI users is approximately:

Tracked locally in your browser — see /progress for the dashboard.