CI Atlas · Hearing Music Through an Implant · Module 05

5Which Instrument Is That? Timbre Perception

Timbre is what lets us tell a violin from a trumpet on the same note. It lives in the fine shape of the harmonic spectrum and in the attack of each tone - and the implant blurs the first while partly sparing the second, leaving recipients struggling to name instruments and often complaining that everything sounds harsh, dull or alike.

FWhat timbre is made of

Timbre is the 'colour' of a sound - the quality that distinguishes two instruments playing the identical pitch and loudness. It has no single physical correlate; it is multidimensional. Its two dominant ingredients are the SPECTRAL ENVELOPE (the relative strengths of the harmonics, which gives an instrument its characteristic vowel-like resonance) and the TEMPORAL ENVELOPE, especially the ONSET or attack (a plucked string starts sharply, a bowed string swells in). Normal-hearing listeners identify common instruments almost perfectly - about 97% in one large study - because both ingredients reach the brain intact. Whether a recipient can name an instrument depends on how faithfully the implant preserves each of these two cues.[2004][2002]

Two instruments playing the same note share the same harmonic positions; what tells them apart is the spectral envelope — the relative strength of each harmonic. The healthy cochlea resolves those harmonics individually, so the two envelopes look (and sound) distinct. Push the same sounds through only ~8–12 effective CI channels and the implant pools energy into a handful of wide bands, flattening both envelopes into similar staircases. With the timbre cue smeared, the instruments converge — which is why instrument identification through an implant is so hard. Toggle the view to compare. Schematic.

TWhy the spectral envelope is blurred

Recognising the spectral envelope requires resolving where the harmonic peaks sit and how tall they are - fine spectral detail. The implant's coarse spectral resolution (a handful of broad, overlapping channels) smears these peaks together, so the distinctive resonance pattern that marks a clarinet versus an oboe is largely lost. The damage is uneven: high-frequency instruments and the string family are recognised and rated worst, because their identity sits in upper harmonics and subtle spectral shape that the array's apical-to-basal map represents poorly. This is why many instruments collapse toward the same percept and recipients report that strings, brass and woodwinds all sound similar.[2002][2008]

Identifying which instrument is playing depends on timbre, the spectral-shape cue the implant degrades most. Cochlear-implant listeners average around 44% correct on single, isolated instruments, against roughly 97% for normal-hearing listeners. Ensembles are harder for everyone, but they punish CI users disproportionately: with several instruments sounding at once, the already-smeared spectral envelopes overlap and become even less separable (~24% here). Tap a bar group, or a legend item to isolate a group. Illustrative.

TThe attack cue that partly survives

Not all of timbre is spectral. The onset - the speed and shape of a note's attack - is carried by the temporal envelope, which the implant preserves comparatively well. This surviving cue lets recipients distinguish a percussive, sharply-attacked instrument (piano, plucked guitar) from a slowly-swelling, sustained one (bowed strings, organ) even when the harmonic detail is gone. Attack cues therefore prop up instrument identification: where two instruments differ mainly in onset, CI users do better; where they differ mainly in spectral fine structure, they do worst. It also explains a partial rehabilitation route - training and instrument choice can exploit the temporal cues that remain rather than the spectral ones that do not.[2004][2014]

An instrument’s identity is carried partly by its attack — the shape of its onset and temporal envelope — and partly by its fine spectral detail. The implant preserves the temporal envelope well but smears the spectrum, so instruments are sorted along the diagonal: those whose recognisability leans on a sharp attack (percussion, plucked and struck strings, piano) land in the lower-right and survive, while sustained, spectrally-defined instruments (bowed strings, clarinet, trumpet) sit upper-left where the lost cue lived and become hard to name. Schematic.

CThe clinical picture: below normal, and 'harsh, dull, alike'

Across studies, CI instrument identification averages far below normal hearing - on the order of ~44% correct versus ~97% for normal-hearing listeners on the same task. Single instruments are easier than ensembles: when several instruments play together, the already-blurred spectra overlap and become almost impossible to separate. A recurrent qualitative complaint is poor sound quality - instruments described as harsh, tinny, dull, mechanical or indistinguishable from one another, and appraisal ratings are lowest for the very instruments (high-pitched, strings) that are hardest to recognise. Hearing-aid users with comparable hearing loss outperform CI users on timbre, a reminder that the deficit is largely the implant's spectral coding, not the listener's brain - and that structured timbre training measurably improves both recognition and appraisal.[2004][2002][2008]

Case 29.5 · Which Instrument Is That? Timbre P

A 12-year-old experienced CI user tells his audiologist he can always tell when a piano is playing and 'kind of' tell a drum, but in his school orchestra recordings 'all the smooth ones sound the same' and the violins 'sound scratchy and angry.' He still enjoys music overall.

Which explanation best accounts for his pattern of preserved and impaired instrument perception?

Self-assessment — Module 53 questions

Question 1

Which two acoustic features most determine an instrument's timbre?

Question 2

Why do CI users find sustained string instruments especially hard to identify?

Question 3

Compared with cochlear implant users, hearing-aid users with similar hearing loss typically:

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