Harvard Dialect Survey · spoken edition

Say It Like You Say It

How this works

60 questions · ~10 minutes

The dialect quiz at mydialect.us asks you to pick answers off a list. This one asks you to talk. You read short lines aloud; where a word is up to you, the line stops and leaves you a gap to fill with whatever you'd actually say.

Everything you say is measured on this device — vowel formants, sibilant spectra, syllable nuclei, voicing — and turned into an IPA reading, which picks your answer for you. Nothing is uploaded except the speech-recognition transcript, which the browser handles.

On iPhone and iPad, use Safari itself. Every iOS browser runs on WebKit, but speech recognition is Safari's own — Chrome, Firefox and in-app browsers don't get it, and neither does a Home Screen shortcut. The 36 pronunciation questions work in any of them, since those are measured from the audio; the 24 word questions need Safari, or a tap. Turn on Settings → General → Keyboard → Enable Dictation, and expect to tap once to start each line: iOS only opens the microphone from a gesture, and hands it to one thing at a time.

01Calibration. Eleven reference words tune the analyser to your own vowel space. Every later judgement is made relative to your vowels, not a General American yardstick.
02Read the lines. One line at a time — tap to begin, and it stops itself when you stop talking.
03Check the verdict. The IPA and the chosen answer appear, and are read back to you. Tap a different chip if it got you wrong.
04Submit. All 60 answers are encoded and handed to the real quiz page, which draws your map.

Calibration

round 1 of 13
spectrogram · 0–5 kHz

Answer sheet

0 of 60

    Your vowel space

    F1 × F2, measured from your calibration