Learn the IPA

Why the IPA exists

Spelling is a poor guide to pronunciation. In English, though, through, tough and cough all end in “ough” and all sound different. Other languages use the same letters for different sounds: the j in Spanish jamón, German ja and English jam are three different sounds.

The IPA fixes this with one rule: one symbol, one sound, every time. A symbol never changes its value with the word or the language, and a sound is always written with the same symbol. Dictionaries, language textbooks, linguists, singers and speech therapists all use it, and once you can read it you can pronounce words in languages you’ve never studied.

Most of the symbols are ordinary Latin letters. Many keep the value you’d expect (p, b, t, d, k, m, n, f, v, s, z, h, l, w sound much as they do in English). The rest are modified letters (ŋ ʃ ʒ ɹ ɾ), Greek letters (θ β χ) or new shapes (ʔ ʕ).

How consonants are described

Every consonant is named by answering three questions. The name is the description: once you know the three parts, you can work out how to make a sound you’ve never heard.

1. Voicing: do your vocal cords vibrate? Put your fingers on your throat and say a long “sss”, then “zzz”. The mouth is identical; the buzz you feel in /z/ is voicing. /s/ is voiceless, /z/ is voiced. Many consonants come in pairs like this: /p b/, /t d/, /k ɡ/, /f v/, /θ ð/, /ʃ ʒ/.
2. Place: where is the airflow blocked? From the lips back: bilabial (both lips, /p m/), labiodental (lip and teeth, /f/), dental (tongue at the teeth, /θ/), alveolar (the ridge behind the teeth, /t s/), postalveolar (just behind it, /ʃ/), palatal (hard palate, /j/), velar (soft palate, /k ŋ/), uvular, pharyngeal and glottal (the throat itself, /h ʔ/).
3. Manner: how is it blocked? A plosive stops the air completely, then releases it (/p t k/). A nasal stops the mouth but lets air out the nose (/m n ŋ/). A fricative squeezes the air through a narrow gap so it hisses (/f s ʃ/). An approximant barely narrows the gap (/j w ɹ/). Trills and taps flick the tongue; lateral sounds let air round the sides of the tongue (/l/).

Put together, /ʃ/ is a voiceless postalveolar fricative: no buzz, tongue just behind the gum ridge, air hissing through a narrow gap. That’s the “sh” in ship.

How vowels are described

Vowels don’t block the air at all. They differ in the shape of the mouth, described in three ways:

• Height: how high the tongue is. Say “ee” then “ah” and feel your jaw drop. /i/ is close (high), /a/ is open (low).
• Backness: whether the tongue is pushed forward or pulled back. Glide from “ee” to “oo”: the tongue slides from front /i/ to back /u/.
• Rounding: whether the lips are rounded. Say “ee”, then keep your tongue still and round your lips: you get French /y/, as in tu.

Two vowels in a row in one syllable form a diphthong, written with both symbols: /aɪ/ in my, /aʊ/ in now. The most common vowel in English is the schwa /ə/, the relaxed “uh” in unstressed syllables like the a in about. It’s spelled with every vowel letter, which is a big reason English spelling is so hard.

The vowels of American English

• ifleece, see
• ɪkit, sit
• eɪface, day
• ɛdress, bed
• ætrap, cat
• ɑfather, palm
• ɔthought, law
• oʊgoat, no
• ʊfoot, put
• ugoose, food
• ʌstrut, cup
• əabout, sofa
• ɝ / ɚnurse / letter
• aɪprice, my
• aʊmouth, now
• ɔɪchoice, boy

Stress, length and other marks

• ˈ goes before the stressed syllable: record the noun is /ˈɹɛkɚd/, the verb is /ɹɪˈkɔɹd/. ˌ marks a weaker, secondary stress.
• ː makes the sound before it long: British seat /siːt/.
• . separates syllables when it helps: /ˈhæp.i/.
• Small raised letters and marks are diacritics that fine-tune a sound: ʰ means a puff of air, as in [pʰɪn].
• /slashes/ give a broad transcription: just the sounds that change meaning. [brackets] give a narrow one with every detail you can hear. Dictionaries use slashes.

Reading a word in IPA

Sounding out a transcription works the same way every time:

1. Forget the spelling. Read each symbol by its IPA value, not as an English letter. The traps: /j/ is the y in yes, /c/ is never an s sound, /ɡ/ is always hard, /e/ is not “ee” and /i/ is.
2. Find the stress. Look for ˈ and remember that the syllable right after it is the loud one.
3. Go left to right, one symbol at a time. Each symbol is one sound. A tie bar (t͡ʃ) or a diphthong (aɪ) joins two symbols into a single smooth sound.
4. Blend. Say the sounds slowly, then faster until they run together into a word.

Here are some words taken apart sound by sound:

Writing a word in IPA

Building your own transcription is the reverse. It’s a skill of listening, not spelling:

1. Say the word out loud, naturally. Say it the way you’d say it in a sentence, not slowly and carefully. Careful speech adds vowels you don’t normally use.
2. Count the syllables and find the stressed one. banana has three, with stress on the second: ba-NA-na. Put ˈ in front of it.
3. Go sound by sound, not letter by letter. Ask “what am I doing with my mouth?” for each sound. Two letters can be one sound (sh → /ʃ/, ng → /ŋ/), one letter can be two (x → /ks/), and some letters are silent.
4. Check the unstressed vowels. Most of them shrink to /ə/ or /ɪ/: banana is /bəˈnænə/, not /baˈnana/.
5. Check the usual suspects. Is that “th” voiced (/ð/ in this) or voiceless (/θ/ in thin)? Is that “s” really a /z/? Is a final “ng” just /ŋ/?
6. Read it back. Sound out what you wrote using only the symbols. If it doesn’t sound like the word, find the symbol that’s off.

Common spelling-to-sound traps in English

• ʃ

  sh, ti, ci, s, ch: ship, nation, special, sure, chef

• k

  c, k, ck, ch, q: cat, kit, back, school, queen /kwin/

• z

  z, s, x: zoo, dogs, xylophone

• ə

  any unstressed vowel: about, taken, pencil, lemon, supply

• ɾ

  t or d between vowels (American): water, ladder sound alike

Try it yourself

Write each word in IPA (American pronunciation), then tap to check your answer.

Spelling quiz

Build each word from the symbol tiles, one tile per sound. Diphthongs like /aɪ/ and affricates like /t͡ʃ/ count as one sound and come as one tile. Stress marks aren’t needed. When you check, any sound you got wrong comes with an explanation.