CI Atlas · Audiological Evaluation · Module 13

13Speech-in-noise & ecologically valid testing

The hardest listening in real life happens in noise, and a test in quiet can completely miss it — a person may ace single words in a silent booth and still be lost at every dinner table. Speech-in-noise testing measures the thing that actually matters: the signal-to-noise ratio at which understanding holds, found by letting the noise close in adaptively until the listener is right half the time. The details turn out to be everything. Whether the masker is steady or a babble of voices changes the score, because a normal-hearing listener exploits the brief gaps in babble while a cochlear-implant user largely cannot; where the loudspeakers sit changes it too. And a growing movement toward ecological validity argues that our tests should look more like life — fluctuating, spatial, effortful — than like the quiet booth. This module covers method; the candidacy cutoffs are in the next chapter.

TWhy test in noise

Real-world difficulty is in noise, not quiet. The speech-in-noise SRT is the signal-to-noise ratio (dB SNR) for 50% correct, found adaptively — letting a cochlear-implant user be compared with a normal-hearing listener who tolerates a far more negative SNR.[1994]

CThe adaptive tests

The named tests: HINT (BKB-derived sentences, SNR adapted to 50%), QuickSIN (IEEE sentences in babble, scored as the signal-to-babble ratio), BKB-SIN (difficulty-matched list pairs), and AzBio (multi-talker conversational sentences chosen to avoid the ceiling effects that retired HINT from candidacy).[2004][2008]

CMasker type & geometry

Masker type is a deliberate variable: cochlear-implant users gain little from the dips in fluctuating babble (unlike normal-hearing listeners), so babble depresses their scores far more than steady noise — and babble better mimics restaurants and classrooms. Geometry and standards matter too: a sound-treated room meeting ANSI S3.1, the loudspeaker 1 m at 0° azimuth, head centred, and speech and noise from the same 0° speaker so directional microphones do not inflate the score.[1999]

Two test choices shape the result. Masker type: a normal-hearing listener exploits the brief dips in fluctuating babble, but a cochlear-implant user largely cannot, so babble depresses CI scores far more than steady noise — and babble better mimics restaurants and classrooms. Geometry: separating speech and noise gives spatial release, a big gain for normal hearing but a small one for the implant. Realistic, ecologically valid testing therefore uses babble and controlled spatial scenes (Keidser et al.). Schematic.

CEcological validity & effort

Ecological validity (Keidser et al.) reframes test design: conventional quiet/steady-noise tests predict daily function only partly, which motivates spatially-separated, fluctuating, multi-source scenarios. Datalogging shows most everyday speech reaches the ear below 60 dB SPL, motivating tests that sample soft and distant speech. And listening effort and fatigue — the FUEL framework — are outcome dimensions recognition scores miss, increasingly paired with speech-in-noise testing.[2020][2016]

Two people with identical audiograms can differ enormously in noise, and a pure-tone test cannot reveal it. Speech-in-noise measures (such as the QuickSIN) report an SNR loss — how many extra decibels of signal-over-noise a listener needs compared with a normal-hearing person to understand 50% of speech. Grading it (normal → mild → moderate → severe) translates directly into counselling and technology: directional microphones, remote/FM systems, and realistic expectations — information that matters as much as the audiogram for everyday function. Illustrative; schematic.

Case 10.13 · Fine in quiet, lost in babble

An implant user scores 92% on sentences in quiet but struggles badly in a restaurant. The clinic wants a test that reflects this.

What testing approach captures the problem?

Self-assessment — Module 132 questions

Question 1 · Trainee

What does a speech-in-noise test report?

Question 2 · Clinician

Why does multi-talker babble depress CI users' scores more than steady noise?

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