International Phonetic Alphabet

The International Phonetic Alphabet (IPA) is an alphabetic system of phonetic notation based primarily on the Latin script. It was devised by the International Phonetic Association in the late 19th century as a standardized representation of speech sounds in written form.[1] The IPA is used by lexicographers, foreign language students and teachers, linguists, speech–language pathologists, singers, actors, constructed language creators and translators.[2][3]

The IPA is designed to represent those qualities of speech that are part of lexical (and to a limited extent prosodic) sounds in oral language: phones, phonemes, intonation and the separation of words and syllables.[1] To represent additional qualities of speech, such as tooth gnashing, lisping, and sounds made with a cleft lip and cleft palate, an extended set of symbols, the extensions to the International Phonetic Alphabet, may be used.[2]

IPA symbols are composed of one or more elements of two basic types, letters and diacritics. For example, the sound of the English letter ⟨t⟩ may be transcribed in IPA with a single letter, [t], or with a letter plus diacritics, [t̺ʰ], depending on how precise one wishes to be.[note 1] Slashes are used to signal phonemic transcription; thus /t/ is more abstract than either [t̺ʰ] or [t] and might refer to either, depending on the context and language.

Occasionally letters or diacritics are added, removed or modified by the International Phonetic Association. As of the most recent change in 2005,[4] there are 107 segmental letters, an indefinitely large number of suprasegmental letters, 44 diacritics (not counting composites) and four extra-lexical prosodic marks in the IPA. Most of these are shown in the current IPA chart, posted below in this article and at the website of the IPA.[5]

In 1886, a group of French and British language teachers, led by the French linguist Paul Passy, formed what would be known from 1897 onwards as the International Phonetic Association (in French, l'Association phonétique internationale).[6] Their original alphabet was based on a spelling reform for English known as the Romic alphabet, but to make it usable for other languages the values of the symbols were allowed to vary from language to language.[7] For example, the sound [ʃ] (the sh in shoe) was originally represented with the letter ⟨c⟩ in English, but with the digraph ⟨ch⟩ in French.[6] In 1888, the alphabet was revised so as to be uniform across languages, thus providing the base for all future revisions.[6][8] The idea of making the IPA was first suggested by Otto Jespersen in a letter to Paul Passy. It was developed by Alexander John Ellis, Henry Sweet, Daniel Jones, and Passy.[9]

Since its creation, the IPA has undergone a number of revisions. After revisions and expansions from the 1890s to the 1940s, the IPA remained primarily unchanged until the Kiel Convention in 1989. A minor revision took place in 1993 with the addition of four letters for mid central vowels[2] and the removal of letters for voiceless implosives.[10] The alphabet was last revised in May 2005 with the addition of a letter for a labiodental flap.[11] Apart from the addition and removal of symbols, changes to the IPA have consisted largely of renaming symbols and categories and in modifying typefaces.[2]

Extensions to the International Phonetic Alphabet for speech pathology (extIPA) were created in 1990 and were officially adopted by the in 1994.[12]

The general principle of the IPA is to provide one letter for each distinctive sound (speech segment).[13] This means that:

The alphabet is designed for transcribing sounds (phones), not phonemes, though it is used for phonemic transcription as well. A few letters that did not indicate specific sounds have been retired (⟨ˇ⟩, once used for the "compound" tone of Swedish and Norwegian, and ⟨ƞ⟩, once used for the moraic nasal of Japanese), though one remains: ⟨ɧ⟩, used for the sj-sound of Swedish. When the IPA is used for phonemic transcription, the letter–sound correspondence can be rather loose. For example, ⟨c⟩ and ⟨ɟ⟩ are used in the IPA Handbook for /t͡ʃ/ and /d͡ʒ/.

Among the symbols of the IPA, 107 letters represent consonants and vowels, 31 diacritics are used to modify these, and 17 additional signs indicate suprasegmental qualities such as length, tone, stress, and intonation.[note 4] These are organized into a chart; the chart displayed here is the official chart as posted at the website of the IPA.

The letters chosen for the IPA are meant to harmonize with the Latin alphabet.[note 5] For this reason, most letters are either Latin or Greek, or modifications thereof. Some letters are neither: for example, the letter denoting the glottal stop, ⟨ʔ⟩, originally had the form of a dotless question mark, and derives from an apostrophe. A few letters, such as that of the voiced pharyngeal fricative, ⟨ʕ⟩, were inspired by other writing systems (in this case, the Arabic letter ⟨‎⟩, ʿayn, via the reversed apostrophe).[10]

The International Phonetic Alphabet is based on the Latin alphabet, using as few non-Latin forms as possible.[6] The Association created the IPA so that the sound values of most consonant letters taken from the Latin alphabet would correspond to "international usage" (approximately Classical Latin).[6] Hence, the letters ⟨b⟩, ⟨d⟩, ⟨f⟩, (hard) ⟨ɡ⟩, (non-silent) ⟨h⟩, (unaspirated) ⟨k⟩, ⟨l⟩, ⟨m⟩, ⟨n⟩, (unaspirated) ⟨p⟩, (voiceless) ⟨s⟩, (unaspirated) ⟨t⟩, ⟨v⟩, ⟨w⟩, and ⟨z⟩ have the values used in English; and the vowel letters from the Latin alphabet (⟨a⟩, ⟨e⟩, ⟨i⟩, ⟨o⟩, ⟨u⟩) correspond to the (long) sound values of Latin: [i] is like the vowel in machine, [u] is as in rule, etc. Other letters may differ from English, but are used with these values in other European languages, such as ⟨j⟩, ⟨r⟩, and ⟨y⟩.

This inventory was extended by using small-capital and cursive forms, diacritics and rotation. There are also several symbols derived or taken from the Greek alphabet, though the sound values may differ. For example, ⟨ʋ⟩ is a vowel in Greek, but an only indirectly related consonant in the IPA. For most of these, subtly different glyph shapes have been devised for the IPA, namely ⟨ɑ⟩, ⟨⟩, ⟨ɣ⟩, ⟨ɛ⟩, ⟨ɸ⟩, ⟨⟩, and ⟨ʋ⟩, which are encoded in Unicode separately from their parent Greek letters, though one of them – ⟨θ⟩ – is not, while both Latin ⟨⟩, ⟨⟩ and Greek ⟨β⟩, ⟨χ⟩ are in common use.[15]

The sound values of modified Latin letters can often be derived from those of the original letters.[16] For example, letters with a rightward-facing hook at the bottom represent retroflex consonants; and small capital letters usually represent uvular consonants. Apart from the fact that certain kinds of modification to the shape of a letter generally correspond to certain kinds of modification to the sound represented, there is no way to deduce the sound represented by a symbol from its shape (as for example in Visible Speech) nor even any systematic relation between signs and the sounds they represent (as in Hangul).

Beyond the letters themselves, there are a variety of secondary symbols which aid in transcription. Diacritic marks can be combined with IPA letters to transcribe modified phonetic values or secondary articulations. There are also special symbols for suprasegmental features such as stress and tone that are often employed.

There are two principal types of brackets used to set off (delimit) IPA transcriptions:

All three of the above are provided by the IPA Handbook. The following are not, but may be seen in IPA transcription or in associated material (especially angle brackets):

In most English accents, the phoneme /l/, which is usually spelled as ⟨l⟩ or ⟨ll⟩, is articulated as two distinct allophones: the clear [l] occurs before vowels and the consonant /j/, whereas the dark [ɫ]/[lˠ] occurs before consonants, except /j/, and at the end of words.[27]

IPA letters have cursive forms designed for use in manuscripts and when taking field notes, but the 1999 Handbook of the International Phonetic Association recommended against their use, as cursive IPA is "harder for most people to decipher."[28]

In the early stages of the alphabet, the typographic variants of g, opentail ⟨ɡ⟩ (Opentail g.svg) and looptail g (Looptail g.svg), represented different values, but are now regarded as equivalents. Opentail ⟨ɡ⟩ has always represented a voiced velar plosive, while ⟨Looptail g.svg⟩ was distinguished from ⟨ɡ⟩ and represented a voiced velar fricative from 1895 to 1900.[29][30] Subsequently, ⟨ǥ⟩ represented the fricative, until 1931 when it was replaced again by ⟨ɣ⟩.[31]

In 1948, the Council of the Association recognized ⟨ɡ⟩ and ⟨Looptail g.svg⟩ as typographic equivalents,[32] and this decision was reaffirmed in 1993.[33] While the 1949 Principles of the International Phonetic Association recommended the use of ⟨Looptail g.svg⟩ for a velar plosive and ⟨ɡ⟩ for an advanced one for languages where it is preferable to distinguish the two, such as Russian,[34] this practice never caught on.[35] The 1999 Handbook of the International Phonetic Association, the successor to the Principles, abandoned the recommendation and acknowledged both shapes as acceptable variants.[36]

The International Phonetic Alphabet is occasionally modified by the Association. After each modification, the Association provides an updated simplified presentation of the alphabet in the form of a chart. (See History of the IPA.) Not all aspects of the alphabet can be accommodated in a chart of the size published by the IPA. The alveolo-palatal and epiglottal consonants, for example, are not included in the consonant chart for reasons of space rather than of theory (two additional columns would be required, one between the retroflex and palatal columns and the other between the pharyngeal and glottal columns), and the lateral flap would require an additional row for that single consonant, so they are listed instead under the catchall block of "other symbols".[37] The indefinitely large number of tone letters would make a full accounting impractical even on a larger page, and only a few examples are shown, and even the tone diacritics are not complete; the reversed tone letters are not illustrated at all.

The procedure for modifying the alphabet or the chart is to propose the change in the Journal of the IPA. (See, for example, August 2008 on an open central unrounded vowel and August 2011 on central approximants.)[38] Reactions to the proposal may be published in the same or subsequent issues of the Journal (as in August 2009 on the open central vowel).[39] A formal proposal is then put to the Council of the IPA[40] – which is elected by the membership[41] – for further discussion and a formal vote.[42][43]

Nonetheless, many users of the alphabet, including the leadership of the Association itself, deviate from this norm.[44] The Journal of the IPA finds it acceptable to mix IPA and extIPA symbols in consonant charts in their articles. (For instance, including the extIPA letter 𝼆, rather than ⟨ʎ̝̊⟩, in an illustration of the IPA.)[45]

Of more than 160 IPA symbols, relatively few will be used to transcribe speech in any one language, with various levels of precision. A precise phonetic transcription, in which sounds are specified in detail, is known as a narrow transcription. A coarser transcription with less detail is called a broad transcription. Both are relative terms, and both are generally enclosed in square brackets.[1] Broad phonetic transcriptions may restrict themselves to easily heard details, or only to details that are relevant to the discussion at hand, and may differ little if at all from phonemic transcriptions, but they make no theoretical claim that all the distinctions transcribed are necessarily meaningful in the language.

Phonetic transcriptions of the word international in two English dialects

For example, the English word little may be transcribed broadly as [ˈlɪtəl], approximately describing many pronunciations. A narrower transcription may focus on individual or dialectical details: [ˈɫɪɾɫ] in General American, [ˈlɪʔo] in Cockney, or [ˈɫɪːɫ] in Southern US English.

Phonemic transcriptions, which express the conceptual counterparts of spoken sounds, are usually enclosed in slashes (/ /) and tend to use simpler letters with few diacritics. The choice of IPA letters may reflect theoretical claims of how speakers conceptualize sounds as phonemes or they may be merely a convenience for typesetting. Phonemic approximations between slashes do not have absolute sound values. For instance, in English, either the vowel of pick or the vowel of peak may be transcribed as /i/, so that pick, peak would be transcribed as /ˈpik, ˈpiːk/ or as /ˈpɪk, ˈpik/; and neither is identical to the vowel of the French pique which would also be transcribed /pik/. By contrast, a narrow phonetic transcription of pick, peak, pique could be: [pʰɪk], [pʰiːk], [pikʲ].

IPA is popular for transcription by linguists. Some American linguists, however, use a mix of IPA with Americanist phonetic notation or use some nonstandard symbols for various reasons.[46] Authors who employ such nonstandard use are encouraged to include a chart or other explanation of their choices, which is good practice in general, as linguists differ in their understanding of the exact meaning of IPA symbols and common conventions change over time.

Many British dictionaries, including the Oxford English Dictionary and some learner's dictionaries such as the Oxford Advanced Learner's Dictionary and the Cambridge Advanced Learner's Dictionary, now use the International Phonetic Alphabet to represent the pronunciation of words.[47] However, most American (and some British) volumes use one of a variety of pronunciation respelling systems, intended to be more comfortable for readers of English. For example, the respelling systems in many American dictionaries (such as Merriam-Webster) use ⟨y⟩ for IPA [j] and ⟨sh⟩ for IPA [ʃ], reflecting common representations of those sounds in written English,[48] using only letters of the English Roman alphabet and variations of them. (In IPA, [y] represents the sound of the French ⟨u⟩ (as in tu), and [sh] represents the pair of sounds in grasshopper.)

The IPA is also not universal among dictionaries in languages other than English. Monolingual dictionaries of languages with phonemic orthographies generally do not bother with indicating the pronunciation of most words, and tend to use respelling systems for words with unexpected pronunciations. Dictionaries produced in Israel use the IPA rarely and sometimes use the Hebrew alphabet for transcription of foreign words.[49] Bilingual dictionaries that translate from foreign languages into Russian usually employ the IPA, but monolingual Russian dictionaries occasionally use pronunciation respelling for foreign words.[50] The IPA is more common in bilingual dictionaries, but there are exceptions here too. Mass-market bilingual Czech dictionaries, for instance, tend to use the IPA only for sounds not found in Czech.[51]

IPA letters have been incorporated into the alphabets of various languages, notably via the Africa Alphabet in many sub-Saharan languages such as Hausa, Fula, Akan, Gbe languages, Manding languages, Lingala, etc. This has created the need for capital variants. For example, Kabiyè of northern Togo has Ɖ ɖ, Ŋ ŋ, Ɣ ɣ, Ɔ ɔ, Ɛ ɛ, Ʋ ʋ. These, and others, are supported by Unicode, but appear in Latin ranges other than the IPA extensions.

In the IPA itself, however, only lower-case letters are used. The 1949 edition of the IPA handbook indicated that an asterisk ⟨*⟩ may be prefixed to indicate that a word is a proper name,[52] but this convention was not included in the 1999 Handbook.

The IPA has widespread use among classical singers during preparation as they are frequently required to sing in a variety of foreign languages. They are also taught by vocal coaches to perfect diction and improve tone quality and tuning.[53] Opera librettos are authoritatively transcribed in IPA, such as Nico Castel's volumes[54] and Timothy Cheek's book Singing in Czech.[55] Opera singers' ability to read IPA was used by the site Visual Thesaurus, which employed several opera singers "to make recordings for the 150,000 words and phrases in VT's lexical database ... for their vocal stamina, attention to the details of enunciation, and most of all, knowledge of IPA".[56]

Each character, letter or diacritic, is assigned a number, to prevent confusion between similar characters (such as ɵ and θ, ɤ and ɣ, or ʃ and ʄ) in such situations as the printing of manuscripts. The categories of sounds are assigned different ranges of numbers.[57]

The International Phonetic Association organizes the letters of the IPA into three categories: pulmonic consonants, non-pulmonic consonants, and vowels.[58][59]

Pulmonic consonant letters are arranged singly or in pairs of voiceless (tenuis) and voiced sounds, with these then grouped in columns from front (labial) sounds on the left to back (glottal) sounds on the right. In official publications by the IPA, two columns are omitted to save space, with the letters listed among 'other symbols',[60] and with the remaining consonants arranged in rows from full closure (occlusives: stops and nasals), to brief closure (vibrants: trills and taps), to partial closure (fricatives) and minimal closure (approximants), again with a row left out to save space. In the table below, a slightly different arrangement is made: All pulmonic consonants are included in the pulmonic-consonant table, and the vibrants and laterals are separated out so that the rows reflect the common lenition pathway of stop → fricative → approximant, as well as the fact that several letters pull double duty as both fricative and approximant; affricates may be created by joining stops and fricatives from adjacent cells. Shaded cells represent articulations that are judged to be impossible.

Vowel letters are also grouped in pairs—of unrounded and rounded vowel sounds—with these pairs also arranged from front on the left to back on the right, and from maximal closure at top to minimal closure at bottom. No vowel letters are omitted from the chart, though in the past some of the mid central vowels were listed among the 'other symbols'.

A pulmonic consonant is a consonant made by obstructing the glottis (the space between the vocal cords) or oral cavity (the mouth) and either simultaneously or subsequently letting out air from the lungs. Pulmonic consonants make up the majority of consonants in the IPA, as well as in human language. All consonants in English fall into this category.[61]

The pulmonic consonant table, which includes most consonants, is arranged in rows that designate manner of articulation, meaning how the consonant is produced, and columns that designate place of articulation, meaning where in the vocal tract the consonant is produced. The main chart includes only consonants with a single place of articulation.

Non-pulmonic consonants are sounds whose airflow is not dependent on the lungs. These include clicks (found in the Khoisan languages and some neighboring Bantu languages of Africa), implosives (found in languages such as Sindhi, Hausa, Swahili and Vietnamese), and ejectives (found in many Amerindian and Caucasian languages).

Affricates and co-articulated stops are represented by two letters joined by a tie bar, either above or below the letters.[66] Affricates are optionally represented by ligatures (e.g. ʦ, ʣ, ʧ, ʤ, ʨ, ʥ, ꭧ, ꭦ), though this is no longer official IPA usage[1] because a great number of ligatures would be required to represent all affricates this way. Alternatively, a superscript notation for a consonant release is sometimes used to transcribe affricates, for example for t͡s, paralleling ~ k͡x. The letters for the palatal plosives c and ɟ are often used as a convenience for t͡ʃ and d͡ʒ or similar affricates, even in official IPA publications, so they must be interpreted with care.

Co-articulated consonants are sounds that involve two simultaneous places of articulation (are pronounced using two parts of the vocal tract). In English, the [w] in "went" is a coarticulated consonant, being pronounced by rounding the lips and raising the back of the tongue. Similar sounds are [ʍ] and [ɥ]. In some languages, plosives can be double-articulated, for example in the name of Laurent Gbagbo.

Tongue positions of cardinal front vowels, with highest point indicated. The position of the highest point is used to determine vowel height and backness.

The IPA defines a vowel as a sound which occurs at a syllable center.[68] Below is a chart depicting the vowels of the IPA. The IPA maps the vowels according to the position of the tongue.

The vertical axis of the chart is mapped by vowel height. Vowels pronounced with the tongue lowered are at the bottom, and vowels pronounced with the tongue raised are at the top. For example, [ɑ] (the first vowel in father) is at the bottom because the tongue is lowered in this position. [i] (the vowel in "meet") is at the top because the sound is said with the tongue raised to the roof of the mouth.

In a similar fashion, the horizontal axis of the chart is determined by vowel backness. Vowels with the tongue moved towards the front of the mouth (such as [ɛ], the vowel in "met") are to the left in the chart, while those in which it is moved to the back (such as [ʌ], the vowel in "but") are placed to the right in the chart.

In places where vowels are paired, the right represents a rounded vowel (in which the lips are rounded) while the left is its unrounded counterpart.

Diphthongs are typically specified with a non-syllabic diacritic, as in ⟨uɪ̯⟩ or ⟨u̯ɪ⟩, or with a superscript for the on- or off-glide, as in ⟨uᶦ⟩ or ⟨ᵘɪ⟩. Sometimes a tie bar is used: ⟨u͡ɪ⟩, especially if it is difficult to tell if the diphthong is characterized by an on-glide, an off-glide or is variable.

Diacritics are used for phonetic detail. They are added to IPA letters to indicate a modification or specification of that letter's normal pronunciation.[69]

By being made superscript, any IPA letter may function as a diacritic, conferring elements of its articulation to the base letter. Those superscript letters listed below are specifically provided for by the IPA Handbook; other uses can be illustrated with ⟨⟩ ([t] with fricative release), ⟨ᵗs⟩ ([s] with affricate onset), ⟨ⁿd⟩ (prenasalized [d]), ⟨⟩ ([b] with breathy voice), ⟨⟩ (glottalized [m]), ⟨sᶴ⟩ ([s] with a flavor of [ʃ]), ⟨oᶷ⟩ ([o] with diphthongization), ⟨ɯᵝ⟩ (compressed [ɯ]). Superscript diacritics placed after a letter are ambiguous between simultaneous modification of the sound and phonetic detail at the end of the sound. For example, labialized ⟨⟩ may mean either simultaneous [k] and [w] or else [k] with a labialized release. Superscript diacritics placed before a letter, on the other hand, normally indicate a modification of the onset of the sound (⟨⟩ glottalized [m], ⟨ˀm[m] with a glottal onset). (See #Superscript IPA.)

Subdiacritics (diacritics normally placed below a letter) may be moved above a letter to avoid conflict with a descender, as in voiceless ⟨ŋ̊⟩.[69] The raising and lowering diacritics have optional spacing forms ⟨˔⟩, ⟨˕⟩ that avoid descenders.

The state of the glottis can be finely transcribed with diacritics. A series of alveolar plosives ranging from open-glottis to closed-glottis phonation is:

Additional diacritics are provided by the Extensions to the IPA for speech pathology.

These symbols describe the features of a language above the level of individual consonants and vowels, that is, at the level of syllable, word or phrase. These include prosody, pitch, length, stress, intensity, tone and gemination of the sounds of a language, as well as the rhythm and intonation of speech.[71] Various ligatures of pitch/tone letters and diacritics are provided for by the Kiel convention and used in the IPA Handbook despite not being found in the summary of the IPA alphabet found on the one-page chart.

Under capital letters below we will see how a carrier letter may be used to indicate suprasegmental features such as labialization or nasalization. Some authors omit the carrier letter, for e.g. suffixed [kʰuˣt̪s̟]ʷ or prefixed [ʷkʰuˣt̪s̟],[72] or place a spacing diacritic such as ⟨˔⟩ at the beginning of a word to indicate that the quality applies to the entire word.[73]

Officially, the stress marksˈ ˌ⟩ appear before the stressed syllable, and thus mark the syllable boundary as well as stress (though the syllable boundary may still be explicitly marked with a period).[77] Occasionally the stress mark is placed immediately before the nucleus of the syllable, after any consonantal onset.[78] In such transcriptions, the stress mark does not mark a syllable boundary. The primary stress mark may be doubledˈˈ⟩ for extra stress (such as prosodic stress). The secondary stress mark is sometimes seen doubled ⟨ˌˌ⟩ for extra-weak stress, but this convention has not been adopted by the IPA.[77] Some dictionaries place both stress marks before a syllable, ⟨¦⟩, to indicate that pronunciations with either primary or secondary stress are heard, though this is not IPA usage.[79]

There are three boundary markers: ⟨.⟩ for a syllable break, ⟨|⟩ for a minor prosodic break and ⟨⟩ for a major prosodic break. The tags 'minor' and 'major' are intentionally ambiguous. Depending on need, 'minor' may vary from a foot break to a break in list-intonation to a continuing–prosodic-unit boundary (equivalent to a comma), and while 'major' is often any intonation break, it may be restricted to a final–prosodic-unit boundary (equivalent to a period). The 'major' symbol may also be doubled, ⟨‖‖⟩, for a stronger break.[note 8]

Although not part of the IPA, the following additional boundary markers are often used in conjunction with the IPA: ⟨μ⟩ for a mora or mora boundary, ⟨σ⟩ for a syllable or syllable boundary, ⟨+⟩ for a morpheme boundary, ⟨#⟩ for a word boundary (may be doubled, ⟨##⟩, for e.g. a breath-group boundary),[81]$⟩ for a phrase or intermediate boundary and ⟨%⟩ for a prosodic boundary. For example, C# is a word-final consonant, %V a post-pausa vowel, and T% an IU-final tone (edge tone).

ꜛ ꜜ⟩ are defined in the Handbook as upstep and downstep, concepts from tonal languages. However, the 'upstep' could also be used for pitch reset, and the IPA Handbook illustration for Portuguese uses it for prosody in a non-tonal language.

Phonetic pitch and phonemic tone may be indicated by either diacritics placed over the nucleus of the syllable (e.g. high-pitch ⟨é⟩) or by Chao tone letters placed either before or after the word or syllable. There are three graphic variants of the tone letters: with or without a stave, and facing left or facing right from the stave. The stave was introduced with the 1989 Kiel Convention, as was the option of placing a staved letter after the word or syllable, while retaining the older conventions. There are therefore six ways to transcribe pitch/tone in the IPA: i.e. ⟨é⟩, ⟨˦e⟩, ⟨⟩, ⟨꜓e⟩, ⟨e꜓⟩ and ⟨¯e⟩ for a high pitch/tone.[77][82][83] Of the tone letters, only left-facing staved letters and a few representative combinations are shown in the summary on the Chart, and in practice it is currently more common for tone letters to occur after the syllable/word than before, as in the Chao tradition. Placement before the word is a carry-over from the pre-Kiel IPA convention, as is still the case for the stress and upstep/downstep marks. The IPA endorses the Chao tradition of using the left-facing tone letters, ⟨˥ ˦ ˧ ˨ ˩⟩, for underlying tone, and the right-facing letters, ⟨꜒ ꜓ ꜔ ꜕ ꜖⟩, for surface tone, as occurs in tone sandhi, and for the intonation of non-tonal languages.[84] In the Portuguese illustration in the 1999 Handbook, tone letters are placed before a word or syllable to indicate prosodic pitch (equivalent to [↗︎] global rise and [↘︎] global fall, but allowing more precision), and in the Cantonese illustration they are placed after a word/syllable to indicate lexical tone. Theoretically therefore prosodic pitch and lexical tone could be simultaneously transcribed in a single text, though this is not a formalized distinction.

Rising and falling pitch, as in contour tones, are indicated by combining the pitch diacritics and letters in the table, such as grave plus acute for rising [ě] and acute plus grave for falling [ê]. Only six combinations of two diacritics are supported, and only across three levels (high, mid, low), despite the diacritics supporting five levels of pitch in isolation. The four other explicitly approved rising and falling diacritic combinations are high/mid rising [e᷄], low rising [e᷅], high falling [e᷇], and low/mid falling [e᷆].[85]

The Chao tone letters, on the other hand, may be combined in any pattern, and are therefore used for more complex contours and finer distinctions than the diacritics allow, such as mid-rising [e˨˦], extra-high falling [e˥˦], etc. There are 20 such possibilities. However, in Chao's original proposal, which was adopted by the IPA in 1989, he stipulated that the half-high and half-low letters ⟨˦ ˨⟩ may be combined with each other, but not with the other three tone letters, so as not to create spuriously precise distinctions. With this restriction, there are 8 possibilities.[86]

The old staveless tone letters tend to be more restricted than the staved letters, though not as restricted as the diacritics. Officially, they support as many distinctions as the staved letters,[87] but typically only three pitch levels are distinguished. Unicode supports default or high-pitch ⟨ˉ ˊ ˋ ˆ ˇ ˜ ˙⟩ and low-pitch ⟨ˍ ˏ ˎ ꞈ ˬ ˷⟩. Only a few mid-pitch tones are supported (such as ⟨˗ ˴⟩), and then only accidentally.

Although tone diacritics and tone letters are presented as equivalent on the chart, "this was done only to simplify the layout of the chart. The two sets of symbols are not comparable in this way."[88] Using diacritics, a high tone is ⟨é⟩ and a low tone is ⟨è⟩; in tone letters, these are ⟨⟩ and ⟨⟩. One can double the diacritics for extra-high ⟨⟩ and extra-low ⟨ȅ⟩; there is no parallel to this using tone letters. Instead, tone letters have mid-high ⟨⟩ and mid-low ⟨⟩; again, there is no equivalent among the diacritics.

The correspondence breaks down even further once they start combining. For more complex tones, one may combine three or four tone diacritics in any permutation,[77] though in practice only generic peaking (rising-falling) e᷈ and dipping (falling-rising) e᷉ combinations are used. Chao tone letters are required for finer detail (e˧˥˧, e˩˨˩, e˦˩˧, e˨˩˦, etc.). Although only 10 peaking and dipping tones were proposed in Chao's original, limited set of tone letters, phoneticians often make finer distinctions, and indeed an example is found on the IPA Chart.[89] The system allows the transcription of 112 peaking and dipping pitch contours, including tones that are level for part of their length.

More complex contours are possible. Chao gave an example of [꜔꜒꜖꜔] (mid-high-low-mid) from English prosody.[86]

Chao tone letters generally appear after each syllable, for a language with syllable tone (⟨a˧vɔ˥˩⟩), or after the phonological word, for a language with word tone (⟨avɔ˧˥˩⟩). The IPA gives the option of placing the tone letters before the word or syllable (⟨˧a˥˩vɔ⟩, ⟨˧˥˩avɔ⟩), but this is rare for lexical tone. (And indeed reversed tone letters may be used to clarify that they apply to the following rather than to the preceding syllable: ⟨꜔a꜒꜖vɔ⟩, ⟨꜔꜒꜖avɔ⟩.) The staveless letters are not directly supported by Unicode, but some fonts allow the stave in Chao tone letters to be suppressed.

IPA diacritics may be doubled to indicate an extra degree of the feature indicated.[92] This is a productive process, but apart from extra-high and extra-low tones ⟨ə̋, ə̏⟩ being marked by doubled high- and low-tone diacritics, and the major prosodic break⟩ being marked as a double minor break ⟨|⟩, it is not specifically regulated by the IPA. (Note that transcription marks are similar: double slashes indicate extra (morpho)-phonemic, double square brackets especially precise, and double parentheses especially unintelligible.)

For example, the stress mark may be doubled to indicate an extra degree of stress, such as prosodic stress in English.[93] An example in French, with a single stress mark for normal prosodic stress at the end of each prosodic unit (marked as a minor prosodic break), and a double stress mark for contrastive/emphatic stress:
[ˈˈɑ̃ːˈtre | məˈsjø ‖ ˈˈvwala maˈdam ‖] Entrez monsieur, voilà madame.[94] Similarly, a doubled secondary stress mark ⟨ˌˌ⟩ is commonly used for tertiary (extra-light) stress.[95] In a similar vein, the effectively obsolete (though still official) staveless tone letters were once doubled for an emphatic rising intonation ⟨˶⟩ and an emphatic falling intonation ⟨˵⟩.[96]

Length is commonly extended by repeating the length mark, as in English shhh! [ʃːːː], or for "overlong" segments in Estonian:

(Normally additional degrees of length are handled by the extra-short or half-long diacritic, but the first two words in each of the Estonian examples are analyzed as simply short and long, requiring a different remedy for the final words.)

A number of IPA characters are not consistently used for their official values. A distinction between voiced fricatives and approximants is only partially implemented, for example. Even with the relatively recent addition of the palatal fricative ⟨ʝ⟩ and the velar approximant ⟨ɰ⟩ to the alphabet, other letters, though defined as fricatives, are often ambiguous between fricative and approximant. For forward places, ⟨β⟩ and ⟨ð⟩ can generally be assumed to be fricatives unless they carry a lowering diacritic. Rearward, however, ⟨ʁ⟩ and ⟨ʕ⟩ are perhaps more commonly intended to be approximants even without a lowering diacritic. ⟨h⟩ and ⟨ɦ⟩ are similarly either fricatives or approximants, depending on the language, or even glottal "transitions", without that often being specified in the transcription.

Another common ambiguity is among the palatal consonants. ⟨c⟩ and ⟨ɟ⟩ are not uncommonly used as a typographic convenience for affricates, typically something like [t͜ʃ] and [d͜ʒ], while ⟨ɲ⟩ and ⟨ʎ⟩ are commonly used for palatalized alveolar [n̠ʲ] and [l̠ʲ]. To some extent this may be an effect of analysis, but it is often common for people to match up available letters to the sounds of a language, without overly worrying whether they are phonetically accurate.

It has been argued that the lower-pharyngeal (epiglottal) fricatives ⟨ʜ⟩ and ⟨ʢ⟩ are better characterized as trills, rather than as fricatives that have incidental trilling.[104] This has the advantage of merging the upper-pharyngeal fricatives [ħ, ʕ] together with the epiglottal plosive [ʡ] and trills [ʜ ʢ] into a single pharyngeal column in the consonant chart. However, in Shilha Berber[citation needed] the epiglottal fricatives are not trilled. Although they might be transcribed ⟨ħ̠ ʢ̠⟩ to indicate this, the far more common transcription is ⟨ʜ ʢ⟩, which is therefore ambiguous between languages.

Among vowels, ⟨a⟩ is officially a front vowel, but is more commonly treated as a central vowel. The difference, to the extent it is even possible, is not phonemic in any language.

Three letters are not needed, but are retained due to inertia and would be hard to justify today by the standards of the modern IPA. ⟨ʍ⟩ appears because it is found in English; officially it is a fricative, with terminology dating to the days before 'fricative' and 'approximant' were distinguished. Based on how all other fricatives and approximants are transcribed, one would expect either ⟨⟩ for a fricative (not how it's actually used) or ⟨⟩ for an approximant. Indeed, outside of English transcription, that is what is more commonly found in the literature. ⟨ɱ⟩ is another historic remnant. It is only distinct in a single language, a fact that was discovered after it was standardized in the IPA. A number of consonants without dedicated IPA letters are found in many more languages than that; ⟨ɱ⟩ is retained because of its historical use for European languages, where it could easily be normalized to ⟨⟩. There have been several votes to retire ⟨ɱ⟩ from the IPA, but so far they have failed. Finally, ⟨ɧ⟩ is officially a simultaneous postalveolar and velar fricative, a realization that doesn't appear to exist in any language. It is retained because it is convenient for the transcription of Swedish, where it is used for a consonant that has various realizations in different dialects. That is, it isn't actually a phonetic character at all, but a phonemic one, which is officially beyond the purview of the IPA alphabet.

For all phonetic notation, it is good practice for an author to specify exactly what they mean by the symbols that they use.

Superscript IPA letters may be used to indicate secondary articulation, releases and other transitions, shades of sound, epenthetic and incompletely articulated sounds. In 2020, the International Phonetic Association endorsed the encoding of superscript IPA letters in a proposal to the Unicode Commission for broader coverage of the IPA alphabet. The proposal covered all IPA letters (apart from the tone letters) that were not yet supported, including the implicit retroflex letters ⟨ꞎ 𝼅 𝼈 ᶑ 𝼊⟩, as well as the two length marks ⟨ː ˑ⟩ and old-style affricate ligatures.[45][105] A separate request by the for an expansion of extIPA coverage endorsed superscript variants of all extIPA fricative letters, specifically for the fricative release of consonants.[106] Unicode placed the new superscript ('modifier') letters in a new Latin Extended-F block.

The Unicode characters for superscript (modifier) IPA and extIPA letters are as follows:

The spacing diacritic for ejective consonants, U+2BC, works with superscript letters despite not being superscript itself: ⟨ᵖʼ ᵗʼ ᶜʼ ᵏˣʼ⟩. If a distinction needs to be made, the combining apostrophe U+315 may be used: ⟨ᵖ̕ ᵗ̕ ᶜ̕ ᵏˣ̕⟩. The spacing diacritic should be used for a baseline letter with a superscript release, such as [tˢʼ] or [kˣʼ], where the scope of the apostrophe includes the non-superscript letter, but the combining apostrophe U+315 might be used to indicate a weakly articulated ejective consonant, where the whole consonant is written as a superscript, or together with U+2BC when separate apostrophes have scope over the base and modifier letters, as in ⟨pʼᵏˣ̕⟩.[105]

In addition, the old alternative near-close vowel letters ⟨ɩ⟩ and ⟨ɷ⟩ are supported at U+1DA5 ⟨⟩ and U+107A4 ⟨𐞤⟩. The para-IPA letter for a central reduced vowel, ⟨⟩, is supported at U+1DA7 ⟨⟩; its rounded equivalent, ⟨ᵿ⟩, is not supported by Unicode.

The precomposed rhotic vowel letters ⟨ɚ ɝ⟩ are not supported, as the rhotic diacritic should be used instead: ⟨ᵊ˞ ᶟ˞⟩; similarly with other rhotic vowels.[45]

Superscript length marks can be used, among other things, for indicating the length of aspiration of a consonant, e.g. [pʰ tʰ𐞂 kʰ𐞁]. Another option is to double the diacritic: ⟨kʰʰ⟩.[45]

Superscript letters can be meaningfully modified by combining diacritics, just as baseline letters are. For example, a superscript dental nasal is ⟨ⁿ̪d̪⟩, a superscript voiceless velar nasal is ⟨ᵑ̊ǂ⟩, and labial-velar prenasalization is ⟨ᵑ͡ᵐɡ͡b⟩. Although the diacritic may seem a bit oversized compared to the superscript letter it modifies, as with the composite superscript c-cedilla and the rhotic vowels this can be an aid to legibility: ⟨ᵓ̃⟩.

Spacing diacritics, however, as in ⟨⟩, cannot be secondarily superscripted in plain text: ⟨ᵗʲ⟩.[note 16]

Superscript wildcards are partially supported: e.g. ᴺC (prenasalized consonant), ꟲN (prestopped nasal), Pꟳ (fricative release), CVNᵀ (tone-bearing syllable), Vᴳ (glide/diphthong), Cᴸ liquid/lateral and Cᴿ rhotic/resonant release, NᴾF (epenthetic plosive), Cⱽ (fleeting vowel). However, superscript S and Ʞ for sibilant release and fleeting/epenthetic click release are not supported as of Unicode 14.

A number of IPA letters and diacritics have been retired or replaced over the years. This number includes duplicate symbols, symbols that were replaced due to user preference, and unitary symbols that were rendered with diacritics or digraphs to reduce the inventory of the IPA. The rejected symbols are now considered obsolete, though some are still seen in the literature.

The IPA once had several pairs of duplicate symbols from alternative proposals, but eventually settled on one or the other. An example is the vowel letter ⟨ɷ⟩, rejected in favor of ⟨ʊ⟩. Affricates were once transcribed with ligatures, such as ⟨ʦ ʣ, ʧ ʤ, ʨ ʥ, ꭧ ꭦ⟩ (and others not found in Unicode). These have been officially retired but are still used. Letters for specific combinations of primary and secondary articulation have also been mostly retired, with the idea that such features should be indicated with tie bars or diacritics: ⟨ƍ⟩ for [zʷ] is one. In addition, the rare voiceless implosives, ⟨ƥ ƭ ƈ ƙ ʠ⟩, were dropped soon after their introduction and are now usually written ⟨ɓ̥ ɗ̥ ʄ̊ ɠ̊ ʛ̥⟩. The original set of click letters, ⟨ʇ, ʗ, ʖ, ʞ⟩, was retired but is still sometimes seen, as the current pipe letters ⟨ǀ, ǃ, ǁ, ǂ⟩ can cause problems with legibility, especially when used with brackets ([ ] or / /), the letter ⟨l⟩, or the prosodic marks ⟨|, ‖⟩. (For this reason, some publications which use the current IPA pipe letters disallow IPA brackets.)[107]

Individual non-IPA letters may find their way into publications that otherwise use the standard IPA. This is especially common with:

In addition, it is common to see ad hoc typewriter substitutions, generally capital letters, for when IPA support is not available, e.g. A for ⟨ɑ⟩, B for ⟨β⟩ or ⟨ɓ⟩, D for ⟨ð⟩, ⟨ɗ⟩ or ⟨ɖ⟩, E for ⟨ɛ⟩, F or P for ⟨ɸ⟩, G ⟨ɣ⟩, I ⟨ɪ⟩, L ⟨ɬ⟩, N ⟨ŋ⟩, O ⟨ɔ⟩, S ⟨ʃ⟩, T ⟨θ⟩ or ⟨ʈ⟩, U ⟨ʊ⟩, V ⟨ʋ⟩, X ⟨χ⟩, Z ⟨ʒ⟩, as well as @ for ⟨ə⟩ and 7 or ? for ⟨ʔ⟩. (See also SAMPA and X-SAMPA substitute notation.)

Chart of the Extensions to the International Phonetic Alphabet (extIPA), as of 2015

The "Extensions to the IPA", often abbreviated as "extIPA" and sometimes called "Extended IPA", are symbols whose original purpose was to accurately transcribe disordered speech. At the Kiel Convention in 1989, a group of linguists drew up the initial extensions,[109] which were based on the previous work of the PRDS (Phonetic Representation of Disordered Speech) Group in the early 1980s.[110] The extensions were first published in 1990, then modified, and published again in 1994 in the Journal of the International Phonetic Association, when they were officially adopted by the ICPLA.[111] While the original purpose was to transcribe disordered speech, linguists have used the extensions to designate a number of sounds within standard communication, such as hushing, gnashing teeth, and smacking lips,[2] as well as regular lexical sounds such as lateral fricatives that do not have standard IPA symbols.

In addition to the Extensions to the IPA for disordered speech, there are the conventions of the Voice Quality Symbols, which include a number of symbols for additional airstream mechanisms and secondary articulations in what they call "voice quality".

Capital letters and various characters on the number row of the keyboard are commonly used to extend the alphabet in various ways.

There are various punctuation-like conventions for linguistic transcription that are commonly used together with IPA. Some of the more common are:

⟨**⟩ (a) A reconstructed form, deeper (more ancient) than a single ⟨*⟩, used when reconstructing even further back from already-starred forms.

⟨×⟩ An ungrammatical form. A less common convention than ⟨*⟩ (b), this is sometimes used when reconstructed and ungrammatical forms occur in the same text.

⟨%⟩ A generalized form, such as a typical shape of a wanderwort that has not actually been reconstructed.[113]

⟨$⟩ A phonological word boundary; e.g. ⟨H$⟩ for a high tone that occurs in such a position.

Full capital letters are not used as IPA symbols, except as typewriter substitutes (e.g. N for ⟨ŋ⟩, S for ⟨ʃ⟩, O for ⟨ɔ⟩ – see SAMPA). They are, however, often used in conjunction with the IPA in two cases:

Wildcards are commonly used in phonology to summarize syllable or word shapes, or to show the evolution of classes of sounds. For example, the possible syllable shapes of Mandarin can be abstracted as ranging from /V/ (an atonic vowel) to /CGVNᵀ/ (a consonant-glide-vowel-nasal syllable with tone), and word-final devoicing may be schematized as C/_#. In speech pathology, capital letters represent indeterminate sounds, and may be superscripted to indicate they are weakly articulated: e.g. [ᴰ] is a weak indeterminate alveolar, [ᴷ] a weak indeterminate velar.[114]

⟨V⟩, ⟨F⟩ and ⟨C⟩ have completely different meanings as Voice Quality Symbols, where they stand for "voice" (though generally meaning secondary articulation, as in a 'nasal voice', rather than phonetic voicing), "falsetto" and "creak". They may also take diacritics that indicate what kind of voice quality an utterance has, and may be used to extract a suprasegmental feature that occurs on all susceptible segments in a stretch of IPA. For instance, the transcription of Scottish Gaelic [kʷʰuˣʷt̪ʷs̟ʷ] 'cat' and [kʷʰʉˣʷt͜ʃʷ] 'cats' (Islay dialect) can be made more economical by extracting the suprasegmental labialization of the words: Vʷ[kʰuˣt̪s̟] and Vʷ[kʰʉˣt͜ʃ].[120] The usual wildcard X or C might be used instead of V (i.e., Xʷ[kʰuˣt̪s̟] for all segments labialized, Cʷ[kʰuˣt̪s̟] for all consonants labialized), or omitted altogether (ʷ[kʰuˣt̪s̟]), so that the reader does not misinterpret ⟨⟩ as meaning that only vowels are labialized. (See #Suprasegmentals for other transcription conventions.)

The blank cells on the IPA chart can be filled without too much difficulty if the need arises. Some ad hoc letters have appeared in the literature for the retroflex lateral flap and the retroflex clicks (having the expected forms of ⟨ɺ⟩ and ⟨ǃ⟩ plus a retroflex tail; the analogous ⟨⟩ for a retroflex implosive is even mentioned in the IPA Handbook), the voiceless lateral fricatives (now provided for by the extIPA), the epiglottal trill (arguably covered by the generally trilled epiglottal "fricatives" ⟨ʜ ʢ⟩), the labiodental plosives (⟨ȹ ȸ⟩ in some old Bantuist texts) and the near-close central vowels (⟨ᵻ ᵿ⟩ in some publications). Diacritics can duplicate some of those, such as ⟨ɭ̆⟩ for the lateral flap, ⟨p̪ b̪⟩ for the labiodental plosives and ⟨ɪ̈ ʊ̈⟩ for the central vowels, and are able to fill in most of the remainder of the charts.[121] If a sound cannot be transcribed, an asterisk ⟨*⟩ may be used, either as a letter or as a diacritic (as in ⟨k*⟩ sometimes seen for the Korean "fortis" velar).

Representations of consonant sounds outside of the core set are created by adding diacritics to letters with similar sound values. The Spanish bilabial and dental approximants are commonly written as lowered fricatives, [β̞] and [ð̞] respectively.[122] Similarly, voiced lateral fricatives would be written as raised lateral approximants, [ɭ˔ ʎ̝ ʟ̝]. A few languages such as Banda have a bilabial flap as the preferred allophone of what is elsewhere a labiodental flap. It has been suggested that this be written with the labiodental flap letter and the advanced diacritic, [ⱱ̟].[123]

Similarly, a labiodental trill would be written [ʙ̪] (bilabial trill and the dental sign), and labiodental stops [p̪ b̪] rather than with the ad hoc letters sometimes found in the literature. Other taps can be written as extra-short plosives or laterals, e.g. [ɟ̆ ɢ̆ ʟ̆], though in some cases the diacritic would need to be written below the letter. A retroflex trill can be written as a retracted [r̠], just as non-subapical retroflex fricatives sometimes are. The remaining consonants, the uvular laterals (ʟ̠ etc.) and the palatal trill, while not strictly impossible, are very difficult to pronounce and are unlikely to occur even as allophones in the world's languages.

The vowels are similarly manageable by using diacritics for raising, lowering, fronting, backing, centering, and mid-centering.[124] For example, the unrounded equivalent of [ʊ] can be transcribed as mid-centered [ɯ̽], and the rounded equivalent of [æ] as raised [ɶ̝] or lowered [œ̞] (though for those who conceive of vowel space as a triangle, simple [ɶ] already is the rounded equivalent of [æ]). True mid vowels are lowered [e̞ ø̞ ɘ̞ ɵ̞ ɤ̞ o̞] or raised [ɛ̝ œ̝ ɜ̝ ɞ̝ ʌ̝ ɔ̝], while centered [ɪ̈ ʊ̈] and [ä] (or, less commonly, [ɑ̈]) are near-close and open central vowels, respectively. The only known vowels that cannot be represented in this scheme are vowels with unexpected roundedness, which would require a dedicated diacritic, such as protruded ⟨ʏʷ⟩ and compressed ⟨uᵝ⟩ (or protruded ⟨ɪʷ⟩ and compressed ⟨ɯᶹ⟩).

An IPA symbol is often distinguished from the sound it is intended to represent, since there is not necessarily a one-to-one correspondence between letter and sound in broad transcription, making articulatory descriptions such as "mid front rounded vowel" or "voiced velar stop" unreliable. While the Handbook of the International Phonetic Association states that no official names exist for its symbols, it admits the presence of one or two common names for each.[125] The symbols also have nonce names in the Unicode standard. In many cases, the names in Unicode and the IPA Handbook differ. For example, the Handbook calls ɛ "epsilon", but Unicode calls it "small letter open e".

The traditional names of the Latin and Greek letters are usually used for unmodified letters.[note 17] Letters which are not directly derived from these alphabets, such as [ʕ], may have a variety of names, sometimes based on the appearance of the symbol or on the sound that it represents. In Unicode, some of the letters of Greek origin have Latin forms for use in IPA; the others use the letters from the Greek section.

For diacritics, there are two methods of naming. For traditional diacritics, the IPA notes the name in a well known language; for example, é is e-acute, based on the name of the diacritic in English and French. Non-traditional diacritics are often named after objects they resemble, so is called d-bridge.

Geoffrey Pullum and William Ladusaw list a variety of names in use for IPA symbols, both current and retired, in their Phonetic Symbol Guide.[10]

IPA typeface support is increasing, and nearly complete IPA support with good diacritic rendering is provided by a few typefaces that come pre-installed with various computer operating systems, such as Calibri, as well as some freely available but commercial fonts such as Brill, but most pre-installed fonts, such as the ubiquitous Arial, Noto Sans and Times New Roman, are neither complete nor render many diacritics properly.

Typefaces that provide nearly full IPA support, properly render diacritics and are freely available include:

Free typefaces that provide good IPA support, but don't handle combinations of diacritics or tone letters well, include:

Web browsers generally do not need any configuration to display IPA characters, provided that a typeface capable of doing so is available to the operating system.

Several systems have been developed that map the IPA symbols to ASCII characters. Notable systems include SAMPA and X-SAMPA. The usage of mapping systems in on-line text has to some extent been adopted in the context input methods, allowing convenient keying of IPA characters that would be otherwise unavailable on standard keyboard layouts.

IETF language tags have registered fonipa as a variant subtag identifying text as written in IPA.[126] Thus, an IPA transcription of English could be tagged as en-fonipa. For the use of IPA without attribution to a concrete language, und-fonipa is available.

Online IPA keyboard utilities[127] are available, and they cover the complete range of IPA symbols and diacritics. In April 2019, Google's Gboard for Android added an IPA keyboard to its platform.[128][129] For iOS there are multiple free keyboard layouts available, e.g. "IPA Phonetic Keyboard".[130]