Phonetics
International Phonetic Alphabet (IPA)
One symbol per sound, every sound in every language
The IPA is a system of symbols designed so that each symbol stands for exactly one speech sound and each speech sound is written with exactly one symbol. It is maintained by the International Phonetic Association, founded in 1886, and aims to represent every distinctive sound that occurs in any human language. The current chart contains 107 base letters, 52 diacritics and 4 prosodic marks, organized by place and manner of articulation for consonants and by tongue height and frontness for vowels.
- First published1888
- Latest revision2020
- Base letters107
- Diacritics52
- Phonemic notation/slashes/
- Phonetic notation[brackets]
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why the IPA exists
Every writing system was designed for a particular language at a particular moment, and most have drifted ever since. English spelling fixes pronunciations from before the Great Vowel Shift. French keeps consonants nobody pronounces. Mandarin pinyin uses ⟨q⟩ for [tɕʰ] and ⟨x⟩ for [ɕ]. The same letter ⟨c⟩ is /k/ in Italian, /s/ in French before e, /tʃ/ in English cello, /ts/ in German.
This is fine for native readers, who memorize the conventions. It is hopeless for anything cross-linguistic: comparing dialects, teaching pronunciation, documenting an undocumented language, training speech recognition. Linguists need a notation that decouples sound from spelling, and assigns a stable symbol to every distinct articulation.
The IPA was that project, begun by a group of French and English language teachers led by Paul Passy in 1886. Their stated principle: one letter, one sound. The first chart, published in 1888 in the journal Le Maître phonétique, has been revised eleven times since.
How the chart is organized
The IPA is not a list. It is a grid that encodes how each sound is produced.
- Pulmonic consonants sit in a table whose rows are manner of articulation (the kind of obstruction: stop, nasal, trill, tap, fricative, approximant, lateral) and whose columns are place of articulation (where the obstruction happens: bilabial, labiodental, dental, alveolar, postalveolar, retroflex, palatal, velar, uvular, pharyngeal, glottal).
- Voicing is shown by paired symbols within each cell: voiceless on the left, voiced on the right. Where only one symbol appears, only one voicing is attested.
- Non-pulmonic consonants — clicks, ejectives, implosives — sit in a separate table because they are made with non-lung airstreams.
- Vowels are plotted on a trapezoid that abstracts the oral cavity: high-low corresponds to tongue height, front-back to tongue advancement. Pairs are shown unrounded then rounded.
- Diacritics modify the basic symbols: nasalization [˜], aspiration [ʰ], length [ː], dental [t̪], retracted [a̠], etc.
- Suprasegmentals mark stress (ˈ primary, ˌ secondary), syllable boundaries (.), tone, and intonation.
A slice of the consonant chart
The consonant grid is large. A representative slice (pulmonic stops, nasals, fricatives) shows the structure clearly.
| Manner | Bilabial | Labiodental | Alveolar | Postalveolar | Velar | Glottal |
|---|---|---|---|---|---|---|
| Stop (voiceless / voiced) | p / b | — | t / d | — | k / g | ʔ / — |
| Nasal | m | ɱ | n | — | ŋ | — |
| Fricative (voiceless / voiced) | ɸ / β | f / v | s / z | ʃ / ʒ | x / ɣ | h / ɦ |
| Approximant | — | ʋ | ɹ | — | ɰ | — |
| Lateral approximant | — | — | l | — | ʟ | — |
| Trill | ʙ | — | r | — | — | — |
Empty cells without explanation represent unattested but possible articulations. Greyed-out cells (in the published chart) represent articulations judged impossible — for instance, a velar trill, because the back of the tongue cannot vibrate freely.
The vowel trapezoid
Vowels are continuous in articulatory space, so the chart uses a four-corner abstraction with cardinal reference points.
| Front | Central | Back | |
|---|---|---|---|
| Close (high) | i (unrounded) / y (rounded) | ɨ / ʉ | ɯ / u |
| Close-mid | e / ø | ɘ / ɵ | ɤ / o |
| Mid | — | ə (schwa) | — |
| Open-mid | ɛ / œ | ɜ / ɞ | ʌ / ɔ |
| Open (low) | a / ɶ | ä | ɑ / ɒ |
Daniel Jones's eight cardinal vowels — i, e, ɛ, a, ɑ, ɔ, o, u — anchor the system. Real-language vowels are described by where they sit on this lattice. American English /æ/ is between cardinal /ɛ/ and /a/. French /y/ is the rounded counterpart of /i/; German über's vowel is the same /yː/.
The same word in IPA across languages
To see how IPA flattens orthographic noise, compare the same concept written in five languages and then in IPA.
| Language | Spelling | IPA (broad) | Notes |
|---|---|---|---|
| English | night | /naɪt/ | Diphthong /aɪ/, silent ⟨gh⟩ |
| German | Nacht | /naxt/ | ⟨ch⟩ is /x/ after a back vowel |
| French | nuit | /nɥi/ | Labiopalatal approximant /ɥ/ |
| Spanish | noche | /ˈnotʃe/ | Trochaic stress, /tʃ/ for ⟨ch⟩ |
| Mandarin | 夜 yè | /jɛ˥˩/ | Falling tone written with tone letters |
Five different orthographies become five precise sound-strings, comparable directly. This is why every introductory linguistics course teaches the IPA before any other tool: it is the lingua franca of phonetics and phonology.
Worked example: transcribing "photograph"
Take the English word photograph. Spelling: nine letters, two of them silent depending on accent. IPA tells us what is actually pronounced.
- Identify phonemes. Eight: /f/, /oʊ/, /t/, /ə/, /g/, /ɹ/, /æ/, /f/. Broad transcription: /ˈfoʊtəˌgɹæf/.
- Mark stress. Primary on the first syllable, secondary on the third: ˈfoʊ-tə-ˌgɹæf.
- Add allophonic detail (narrow). The first /f/ is plain [f]; the /t/ is unaspirated because it is in an unstressed syllable, surfacing as [ɾ] in American English: [ˈfoʊɾəˌgɹæf]. The vowel before /n/ would be nasalized — there is no /n/ here, so no nasalization.
- Compare to the related word photography. Stress shifts: /fəˈtɑgɹəfi/. The same lexical root surfaces with completely different vowels because English vowel reduction is stress-driven. Only IPA captures this — orthography hides it.
Competing notations and theoretical debates
The IPA is not the only phonetic alphabet, and within it there are choices that have been argued for over a century.
| System | Origin / use | Strengths | Weaknesses |
|---|---|---|---|
| IPA | International Phonetic Association, 1888 | Universal, articulatorily organized, peer-maintained | Many symbols not on standard keyboards; learners struggle with diacritics |
| Americanist Phonetic Notation (APA) | US fieldwork tradition (Boas, Sapir) | Fewer special characters, easier on typewriters | Inconsistent across authors; ⟨š⟩ for /ʃ/, ⟨č⟩ for /tʃ/ — only partially overlaps with IPA |
| SAMPA / X-SAMPA | 1989, ASCII-only IPA | Typeable on any keyboard; used in speech tech | Read-only mapping to IPA; not standalone |
| Pinyin (Mandarin) | 1958, official Chinese romanization | Optimized for one language; simple | Not phonetic outside Mandarin; ⟨q⟩, ⟨x⟩, ⟨zh⟩ misleading |
| Dictionary respelling | Webster, Oxford and similar | Readable to non-linguists | English-specific; does not transcribe other languages |
| Praat / acoustic feature notation | Lab phonetics | Tied to measurable parameters (F1, F2, VOT) | Not human-readable for dictionary use |
Internal IPA debates include: whether to keep separate symbols for sounds that turn out to be allophonic in nearly all languages (the labiodental flap [ⱱ] was added only in 2005); whether to standardize one symbol per articulation or per phonologically distinct category; how finely the vowel space should be subdivided. The Association votes on changes; revisions are conservative.
Variants and applications
- extIPA. Extensions to the IPA for Disordered Speech, used by speech-language pathologists; adds symbols for sounds that arise in clinical populations (e.g., labioalveolar contact).
- VoQS. Voice Quality Symbols, a parallel set for voice quality (creaky, breathy, harsh, whispery).
- ToBI. Tones and Break Indices, for prosodic transcription of intonation; commonly used alongside IPA.
- SAMPA / X-SAMPA. ASCII-mapped versions of the IPA, designed for speech-recognition databases and email-era linguistics; X-SAMPA is the more complete.
- IPA-Unicode. Most IPA symbols are encoded in Unicode (block U+0250–U+02AF, with extensions). Modern linguistic work uses Unicode IPA fonts like Doulos SIL, Charis SIL, or system fonts.
Common pitfalls
- Confusing slashes and brackets. /p/ is a phoneme; [p] is a phonetic realization. The convention matters; mixing them obscures whether you are claiming contrast or detail.
- Treating IPA as a pronunciation guide for English speakers. ⟨a⟩ in IPA is a low front unrounded vowel, not the English ⟨a⟩ in cat. ⟨e⟩ is mid-front, not the diphthong of English day. Read every symbol as IPA, not as your native orthography.
- Forgetting the diacritics. [t] and [tʰ] differ; so do [a] and [ã]; so do [s] and [s̪]. Narrow transcription is the diacritics.
- Using IPA for orthographic respelling. Some popular guides write English night as [nɪɡʰt]. This is incorrect: it is /naɪt/.
- Assuming the chart is closed. The IPA evolves; recent changes added the labiodental flap (2005) and revised the chart layout (2020). Always check the latest version.
- Confusing tonemes and tone letters. Tone is suprasegmental; the five tone letters (˥ ˦ ˧ ˨ ˩) are stacked to indicate contour, not appended like vowel diacritics.
Frequently asked questions
Why does linguistics need a separate alphabet?
Because writing systems do not transcribe sound. English spelling represents the same sound several ways ("see, sea, scene") and the same letters several sounds ("cough, though, through"). Different languages use the same letters for different sounds (German "w" is /v/). The IPA gives one symbol per sound, regardless of language or orthography, so that linguists can compare languages and capture pronunciation precisely.
What is the difference between broad and narrow transcription?
Broad (phonemic) transcription uses slashes /…/ and lists only the contrastive sounds — the phonemes — of a language, ignoring predictable variation. "Pin" is /pɪn/. Narrow (phonetic) transcription uses square brackets […] and adds the predictable allophonic detail, including aspiration, nasalization, length, and other diacritics: [pʰɪ̃n]. Use broad for general work; use narrow when the phonetic detail is the point.
How is the IPA chart organized?
Consonants are arranged in a grid: rows are manner of articulation (stop, fricative, nasal, etc.), columns are place of articulation (bilabial, alveolar, velar, etc.). Voiceless symbols sit on the left of each cell, voiced on the right. Vowels are arranged on a trapezoid that mirrors the oral cavity: front-back left-to-right, high-low top-to-bottom, with rounded vowels paired to the right of unrounded ones.
Where do IPA symbols come from?
Most are Latin or Greek letters used with their commonest values: ⟨p⟩ is /p/, ⟨t⟩ is /t/, ⟨θ⟩ is the Greek theta. Where Latin and Greek run out, the IPA uses turned and flipped versions (⟨ɐ⟩ is turned ⟨a⟩, ⟨ə⟩ schwa is turned ⟨e⟩), tail-modified letters (⟨ɲ⟩, ⟨ʃ⟩) or wholly novel symbols (⟨ʔ⟩ glottal stop). Each addition is voted on by the International Phonetic Association.
Why are some cells in the chart shaded?
Shaded cells in the IPA chart indicate articulations judged impossible. You cannot make a velar trill, for instance, because the back of the tongue cannot vibrate freely against the velum. Empty cells represent possible but unattested articulations. The boundary moves: the labiodental flap [ⱱ] was added in 2005 after African languages were documented using it.
Is IPA the same as phonemic spelling?
No. IPA is the symbol set; phonemic transcription is one application of it. The IPA can also write narrow phonetic transcription, allophonic detail, and exotic sounds in unanalyzed languages. Some traditions, especially in dictionaries for English learners, use a simplified "respelling" instead of full IPA — but most modern linguistic work uses IPA throughout.