Measuring speech intelligibility means finding out how much of your words people understand. At home, record set words and short phrases, have unfamiliar listeners write what they hear, and score percent correct. Re-test every 2 weeks with the same setup. You can also run your recordings through speech-to-text (ASR) as a proxy. Track the average over time, not a single score.
What does “speech intelligibility” mean?
Speech intelligibility is how much of your spoken message a listener understands. Clinicians often measure it as the percent of words or sentences understood by unfamiliar listeners in a standard task. It is a useful way to track change for a person with dysarthria or other motor speech disorders. See definitions and assessment ideas from ASHA and the NIDCD.
How do I measure intelligibility speech at home?
Use a repeatable, simple plan. Aim for percent-correct scores and a short rating.
- Make your test set
- 50 single words, 10 everyday phrases, and a 1 to 2 minute paragraph.
- Example words: book, water, coffee, window, phone, dinner, jacket, paper, number, music.
- Example phrases: “My name is __.” “Can we meet at 3?” “I need a minute.” “Please repeat that.” “I will email you.”
- Record in a quiet room
- Use the same device each time. Keep the mic 6 to 8 inches from your mouth.
- Say the date. Read the words and phrases once each. Do not spell or explain.
- Recruit unfamiliar listeners
- Ask 3 to 5 people who do not talk with you daily. They can be friends of friends or online volunteers.
- Tell them to write exactly what they hear. No guessing based on context.
- Score percent correct
- Words: Percent = (number correct / total) x 100.
- Phrases: Count content words (nouns, main verbs) correct and score the same way.
- Average across listeners to reduce bias.
- Add a listener rating
- After each clip, have listeners rate overall clarity on a 5-point scale: 1 = hard to understand, 5 = easy to understand.
- Use ASR as a proxy
- Run the same recordings through on-device or offline ASR. Save the transcript. Score percent words correct versus your script.
- ASR is not a medical test, but trends can match human listener results for many speakers.
- Repeat and compare
- Re-test every 2 weeks. Use the same lists, mic, and room.
- Graph the averages. Look for stable change, not one-off spikes.
Tip: Record your phrases once, then relisten in 2 weeks when you or your listeners have forgotten the exact lines. This reduces memory bias.
What lists should I use and how many words?
Keep it practical and repeatable.
- Words: 50 common items is a good start. You can add another 50 if your scores sit at 0 percent or 100 percent.
- Phrases: 10 short, functional phrases you actually say during your day.
- Paragraph: 1 to 2 minutes of reading. Use the same text each time.
If you want standardized materials, an SLP can guide you to tools like the Assessment of Intelligibility of Dysarthric Speech and the Sentence Intelligibility Test, which follow set rules for word and sentence choice. See ASHA’s overview of dysarthria assessment.
How to reduce bias and track real change
Small setup differences can swing scores. Control what you can.
- Same device, mic distance, and room each session.
- Test at the same time of day, especially if fatigue affects your voice.
- Use unfamiliar listeners and average 3 to 5 scores.
- Keep the original scripts private from listeners.
- When comparing to ASR, use the exact same audio files.
- Watch trends across months, not days. A 10-point change that holds over 2 to 3 sessions is more meaningful than a one-time bump.
Can I use ASR to measure intelligibility speech?
Yes, as one data point. If a speech-to-text engine transcribes more of your words, people often understand you better too. But ASR is trained mostly on typical speech. It can miss accented or dysarthric speech, or do better than a human who lacks context. Treat ASR accuracy as a proxy, not a diagnosis. For background on dysarthria and why clarity varies, see the NIDCD.
If privacy matters, test with tools that work offline. DysVoxa is a free, on-device app for speech correction and text-to-speech. It keeps audio on your computer and offers a virtual microphone for calls. Read about the 4-step loop at how DysVoxa works and our privacy-first design.
A simple comparison of methods
| Method | What you measure | How to do it | Pros | Limits |
|---|---|---|---|---|
| Listener transcription | Percent words understood | Unfamiliar listeners write what they hear | Gold standard for real listeners | Needs volunteers and time |
| Listener 5-point rating | Overall clarity | Quick rating after each clip | Fast snapshot | Less precise than percent |
| ASR accuracy | Proxy for intelligibility | Run audio through speech-to-text | Easy to repeat, instant score | Can misjudge atypical speech |
| Paragraph reading | Connected speech clarity | Read 1-2 minute text | Reflects real-life speaking | Reading skill can affect results |
| Formal SLP tests | Standardized measure | SIT/AIDS with an SLP | Reliable, comparable | Needs clinic access, cost |
How DysVoxa can support your tracking
DysVoxa is not therapy or a medical device. It is a free forever desktop app that runs fully offline after install. It combines on-device speech recognition, correction, and text-to-speech. You can:
- Record practice sets, review transcripts, and save scores locally.
- Create personal voice profiles to tune recognition.
- Route corrected speech through a virtual microphone to Zoom or Teams.
- Keep your audio private on your own computer.
See the record-understand-correct-speak loop, watch short demo videos, or request the open beta at get access. For a quick overview of why we built an offline tool, read the homepage.
When to bring in a speech-language pathologist
Home tracking is helpful, but a speech-language pathologist can provide:
- Standardized testing (for example, Sentence Intelligibility Test).
- Strategies for breath, rate, loudness, and articulation.
- Tailored goals that fit your daily life.
- Documentation for clinical care.
Find assessment guidance and referrals through ASHA. If you are a clinician exploring tech supports, see our page for clinicians.
Step-by-step example you can copy today
- Print a script of 50 words and 10 phrases you picked.
- Record once in a quiet room with your phone 6 to 8 inches away.
- Share the audio with 3 unfamiliar listeners and collect their transcriptions.
- Score each listener’s percent correct for words and phrases. Average them.
- Run the same audio through speech-to-text and note its percent correct.
- Save all files with the date. Put the scores in a simple spreadsheet.
- Re-test in 2 weeks. Do not practice the exact list daily. Rely on the 2-week gap to reduce memory effects.
Safety, respect, and honest expectations
- Measuring intelligibility is about access, not blame. Speech changes can come from conditions like stroke, Parkinson’s disease, ALS, or cerebral palsy. See the NIDCD for causes and basics.
- Scores guide decisions. They are not a diagnosis.
- Tech can help, but it is not a cure. DysVoxa corrects and supports; it does not replace an SLP.
What this means for you
You can measure intelligibility speech at home with simple tools: a script, a phone, a few unfamiliar listeners, and optional ASR. Keep the setup the same, wait 2 weeks between checks, and track the average. If you want offline support, try the free DysVoxa beta while you work with your care team.
