Question · Speech recognition
What is a speech recognition threshold in a hearing test?
A speech recognition threshold is the softest level at which a person can repeat or identify half of a list of spoken words in a hearing test, usually two-syllable spondee words played through headphones. It checks the pure-tone hearing test, and it measures a person's hearing, not how accurate speech recognition software is.
Updated October 17, 20264 min read8 cited facts, each traced to its source
A speech recognition threshold (SRT) is the softest level at which you can recognize half of a set of spoken words. In a hearing test, the audiologist plays words through headphones and lowers the loudness until you start to miss them. The level at which you get about half right is the threshold, measured in decibels (dB). The standard definition is the minimum hearing level for speech at which a person can recognize 50% of the speech material 1.
The words are usually spondees: two-syllable words with equal stress on both syllables, such as "baseball" or "hot dog" 2. Older texts call the same measure the speech reception threshold.
The name is easy to confuse with software. A speech recognition threshold measures a listener's hearing. It says nothing about how accurately a dictation app or a transcription tool turns speech into text. For that, the measure is word error rate, covered in our guide to speech recognition accuracy.
The threshold is a hearing level, found with words instead of beeps
A pure-tone test uses beeps at set pitches to find the faintest sound you can hear. The speech threshold does the same job with real words, so it checks the two against each other. Clinically, the primary purpose of a speech threshold is to serve as a validity check for the pure tone audiogram 3.
The test runs in steps. The audiologist first makes sure you understand the task and the word list, then presents words at falling levels. The threshold is the lowest level at which you still get about half of them right, so it captures the point where speech becomes hard to follow, not the point where you first hear something.
When the speech threshold and the pure-tone average disagree, the test needs a second look
Your pure-tone average (PTA) is a single number from several beep frequencies. If the speech threshold and the PTA are close, the results agree and you can trust both. If they are far apart, the guideline treats the gap as a warning sign. Disagreement between the speech recognition threshold and pure tone average is an indication of inconsistency in test results, and it may be an early sign of pseudohypacusis, meaning exaggerated or non-organic hearing loss 4.
A gap does not prove anything about you. The same guideline lists other causes, including faulty equipment, a patient who misunderstood the instructions, developmental level and irregular hearing sensitivity. The usual response is to retest, not to diagnose from one result.
Sloping losses and some patients change how the number should be read
Sloping hearing loss. Hearing that is worse at high pitches than at low ones is called a sloping loss. For most people, the speech threshold agrees well with the three-frequency average. With a sloping loss, the guideline says agreement may be better with the two-frequency pure tone average 5. Ask which average your audiologist compared against.
Young children and some patients. The standard procedure works for most adults, but certain individuals such as young children and mentally disabled, uncooperative, or neurologically disabled individuals may require modification of the procedure 6. The guideline includes a picture-based spondaic word list for children. Any change must be recorded with the result.
The threshold is not the level for everyday listening
The speech threshold is measured at the softest level you can handle, so it does not show how well you follow conversation at a normal volume. That is the job of a word recognition test. Clinicians usually set its level a fixed number of decibels above the threshold. The authors of a 2009 study write that the most popular approach is to set the presentation level at a particular sensation level (SL) above the speech recognition threshold 7.
That level is not settled for every patient. The same study of 40 adults with sensorineural hearing loss found that for people with moderately severe or severe losses, the best phoneme recognition scores came when words were presented 5 dB below the level at which sound becomes uncomfortably loud 8. The authors note that methods for choosing this level are not well established for a wide range of hearing losses, so one rule does not fit everyone.
What to do with a speech threshold from your own hearing test
- Ask for the number and the word list. A speech threshold is only comparable when the words and the procedure are known. The guideline asks clinicians to record any change.
- Compare it with your pure-tone average. If the two disagree, ask whether the test will be repeated before the results are used.
- Ask about a word recognition score too, if you want to know how well you follow speech, not only how quietly you can hear it.
- Talk to a qualified audiologist about what the numbers mean for you. This page explains the measure in general and is not medical advice.
Related questions and terms
If you are reading about speech recognition software rather than hearing tests, these pages may help: Is word error rate the same as accuracy?, Is speech recognition biased? and Why does speech recognition get names and jargon wrong?. For the software sense of the phrase, see automatic speech recognition.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.The American Speech-Language-Hearing Association recommends spondaic words as the usual test material for the speech recognition threshold, and defines them as two-syllable words with equal stress on both syllables.
American Speech-Language-Hearing Association, American Speech-Language-Hearing Association, 1988 · Original
- 3.
- 4.The American Speech-Language-Hearing Association says that when the speech recognition threshold and the pure tone average disagree, the results are inconsistent, which can be an early sign of pseudohypacusis, or exaggerated hearing loss.
American Speech-Language-Hearing Association, American Speech-Language-Hearing Association, 1988 · Original
- 5.The American Speech-Language-Hearing Association says that for most clinical cases the speech recognition threshold agrees well with the three-frequency pure tone average, but with a sloping loss it may agree better with the two-frequency average.
American Speech-Language-Hearing Association, American Speech-Language-Hearing Association, 1988 · Original
- 6.The American Speech-Language-Hearing Association says the speech threshold procedure may need modifying for young children and for mentally disabled, uncooperative or neurologically disabled people, and that modifications must be recorded.
American Speech-Language-Hearing Association, American Speech-Language-Hearing Association, 1988 · Original
- 7.
- 8.In a study of 40 adults with sensorineural hearing loss, Guthrie and Mackersie found that for people with moderately severe or severe losses, presenting words 5 dB below their uncomfortable loudness level gave the highest phoneme recognition scores.
Leslie A. Guthrie and Carol L. Mackersie, Journal of the American Academy of Audiology, 2009 · Original
Go deeper
Related questions
- Is speech recognition less accurate for some people?
- Is speech recognition more accurate than humans?
- Why does speech recognition get names and jargon wrong?
- Why does Whisper make up words nobody said?
- Can an ESP32 or microcontroller do speech recognition?
- Can large language models do speech recognition?
- Can Whisper AI do text to speech?
- Can Whisper identify different speakers?