Cochlear Implant Atlas
CI Atlas · The Measure of Success: Speech, Hearing and Real-World Outcomes · Module 02

2The Outcome Test Battery

If outcome is plural, measuring it demands a battery rather than a single test. Modern practice pairs open-set monosyllables with open-set sentences, tests in quiet and increasingly in noise, and reaches for closed-set tasks when a listener is too young or too new to manage open-set material. The art lies in choosing materials hard enough to avoid the ceiling yet fair enough to reflect real listening.

TOpen-set words and sentences: the core

The CNC monosyllabic word test comprises phonemically balanced 50-word lists spoken by a male talker after a carrier phrase, scored as percent words correct, and a full 50-word list is given per condition. Open-set sentence tests in common use include AzBio, HINT and the Bamford-Kowal-Bench lists, with AzBio favoured because its talker variability and difficulty resist the ceiling that plagues older materials. The contemporary Minimum Speech Test Battery specifies one 50-word CNC list plus AzBio sentences in quiet and in noise, replacing the earlier battery that had relied on HINT sentences in quiet. Materials are presented from a loudspeaker one metre from the listener at 0 degrees azimuth, at 60 dBA, a level chosen to represent everyday conversational speech rather than the raised 70 dB SPL used in many older protocols.[2020][2012][1952]

Build the Minimum Speech Test Battery

CNC words50 monosyllabic wordsAzBio quiet20 sentences at 60 dBAAzBio in noise+5 or +10 dB SNRHINT quietRETIRED from the battery
CNC words: Open-set word recognition — the hardest, most diagnostic material; no sentence context to lean on.

The Minimum Speech Test Battery pairs CNC words (50 monosyllables), AzBio sentences in quiet (20 sentences at 60 dBA) and AzBio in noise (+5 or +10 dB SNR). Each adds difficulty the last cannot capture. HINT-in-quiet was retired because too many recipients hit 100%, a ceiling that masked the very differences the battery exists to reveal. Schematic.

CThe shift toward sentences in noise

Because everyday listening occurs at signal-to-noise ratios of roughly 0 to 10 dB, adding noise to the battery probes function that quiet testing misses entirely. Adaptive speech-in-noise tests express results as the SNR needed for 50% correct: in the BKB-SIN the babble rises in 3 dB steps from +21 dB down to -6 dB, and the SNR for 50% is computed as 23.5 minus the number of key words repeated. Speech and noise are typically delivered from the same loudspeaker, because routing them to 0 and 180 degrees lets directional microphones flatter performance in a way that overstates real-world benefit. Consistency is a clinical safeguard: testing in noise only when a candidate fails in quiet is cherry-picking, and a recipient who struggles at an artificially harsh -5 dB SNR will likely still struggle there after implantation.[2020][2017][2013]

BKB-SIN staircase → SNR-50

21123-6SNR (dB)babble steps down, sentence by sentence →
Key words correct0SNR-50

BKB-SIN presents sentences in babble that grows louder in 3 dB steps, from an easy +21 dB SNR down to a punishing -6 dB. The listener tallies key words correct, and the threshold for 50% understanding is read straight off the formula SNR-50 = 23.5 − (key words correct). A lower SNR-50 means better listening in noise — the metric that best mirrors a noisy restaurant. Illustrative.

TClosed-set tests and sound-field detection

Closed-set tests, where the listener chooses from a fixed set of pictures or words, are used for young children and low-performing adults who cannot yet manage the open-set format. Paediatric batteries climb a graded ladder, from pattern perception through closed-set vowel- and consonant-based word identification, before open-set words and sentences are attempted. Sound-field aided detection (warble-tone) thresholds confirm that the map gives audible access across frequencies; programmes commonly target aided thresholds of 30 dB HL or better, and lower thresholds tend to accompany better word scores. Detection thresholds verify audibility but not intelligibility, so they complement rather than replace speech-perception testing in the battery.[2020][2013][2009]

Ceiling effect: where scores pile up

010203040% of subjects0%100%score bin →
At 100%28%TestHINT quiet

On easy HINT-in-quiet, about 28% of subjects score a perfect 100% — a wall of bars jammed at ceiling that hides any difference between them. Switch to AzBio and only 0.7% top out; switch to CNC words and 0% do, leaving a broad spread that ranks every listener. A test that everyone aces measures nothing — which is why the field moved to harder material. Schematic.

CCeiling effects and choosing the right difficulty

Easy materials saturate: in one 156-recipient study 28% scored a perfect 100% on HINT sentences in quiet, whereas only 0.7% reached 100% on AzBio and none scored 100% on CNC words. A test at ceiling cannot detect improvement or distinguish good from excellent listeners, which is why HINT in quiet was abandoned as a primary candidacy and outcome measure. Difficulty is tuned by task and by SNR: in noise the presentation SNR (commonly +5 or +10 dB for AzBio) is chosen to avoid both floor and ceiling so the score lands in a sensitive range. A defensible battery therefore layers easy and hard tasks, monosyllables and sentences, quiet and noise, so that some measure remains sensitive whatever the recipient's level.[2008][2012][2020]

Case 18.2 · The Outcome Test Battery
An audiologist evaluating a candidate finds the patient scores 96% on HINT sentences in quiet, comfortably above the 60% candidacy threshold, yet the patient reports being unable to follow conversation in any group setting. The audiologist wants a measure that better reflects this difficulty.

Which change to the test battery is most appropriate?

Self-assessment — Module 22 questions
Question 1

In the BKB-SIN test, how is the signal-to-noise ratio for 50% correct derived?

Question 2

Why was HINT sentences in quiet replaced by AzBio in the contemporary outcome battery?

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