In phonology, an allophone (//; from the Greek ἄλλος, állos, "other" and φωνή, phōnē, "voice, sound") is one of a set of multiple possible spoken sounds, or phones, or signs used to pronounce a single phoneme in a particular language. For example, in English, [t] (as in stop [stɒp]) and the aspirated form [tʰ] (as in top [ˈtʰɒp]) are allophones for the phoneme /t/, while these two are considered to be different phonemes in some languages such as Thai and Hindi. On the other hand, in Spanish, [d] (as in dolor [doˈloɾ]) and [ð] (as in nada [ˈnaða]) are allophones for the phoneme /d/, while these two are considered to be different phonemes in English.
The specific allophone selected in a given situation is often predictable from the phonetic context, with such allophones being called positional variants, but some allophones occur in free variation. Replacing a sound by another allophone of the same phoneme usually does not change the meaning of a word, but the result may sound non-native or even unintelligible.
Native speakers of a given language perceive one phoneme in the language as a single distinctive sound and are "both unaware of and even shocked by" the allophone variations that are used to pronounce single phonemes.
History of concept
The term "allophone" was coined by Benjamin Lee Whorf circa 1929. In doing so, he placed a cornerstone in consolidating early phoneme theory. The term was popularized by George L. Trager and Bernard Bloch in a 1941 paper on English phonology and went on to become part of standard usage within the American structuralist tradition.
Complementary and free-variant allophones and assimilation
Whenever a user's speech is vocalized for a given phoneme, it is slightly different from other utterances, even for the same speaker. That has led to some debate over how real and how universal phonemes really are (see phoneme for details). Only some of the variation is significant, by being detectable or perceivable, to speakers.
There are two types of allophones, based on whether a phoneme must be pronounced using a specific allophone in a specific situation or whether the speaker has the unconscious freedom to choose the allophone that is used.
If a specific allophone from a set of allophones that correspond to a phoneme must be selected in a given context, and using a different allophone for a phoneme would cause confusion or make the speaker sound non-native, the allophones are said to be complementary. The allophones then complement each other, and one of them is not used in a situation in which the usage of another is standard. For complementary allophones, each allophone is used in a specific phonetic context and may be involved in a phonological process.
In other cases, the speaker can freely select from free variant allophones on personal habit or preference, but free variant allophones are still selected in the specific context, not the other way around.
Another example of an allophone is assimilation, in which a phoneme is to sound more like another phoneme. One example of assimilation is consonant voicing and devoicing, in which voiceless consonants are voiced before and after voiced consonants, and voiced consonants are devoiced before and after voiceless consonants.
There are many allophonic processes in English: lack of plosion, nasal plosion, partial devoicing of sonorants, complete devoicing of sonorants, partial devoicing of obstruents, lengthening and shortening vowels, and retraction.
- Aspiration: In English, a voiceless plosive /p, t, k/ is aspirated (has a string explosion of breath) if it is at the beginning of the first or a stressed syllable in a word. For example, [pʰ] as in pin and [p] as in spin are allophones for the phoneme /p/ because they cannot distinguish words (in fact, they occur in complementary distribution). English-speakers treat them as the same sound, but they are different: the first is aspirated and the second is unaspirated (plain). Many languages treat the two phones differently.
- Nasal plosion – In English, a plosive (/p, t, k, b, d, ɡ/) has nasal plosion if it is followed by a nasal, whether within a word or across a word boundary.
- Partial devoicing of sonorants: In English, sonorants (/j, w, l, r, m, n, ŋ/) are partially devoiced after a voiceless sound in the same syllable.
- Complete devoicing of sonorants: In English, a sonorant is completely devoiced after an aspirated plosive (/p, t, k/).
- Partial devoicing of obstruents: In English, a voiced obstruent is partially devoiced next to a pause or next to a voiceless sound within a word or across a word boundary.
- Retraction: In English, /t, d, n, l/ are retracted before /r/.
Because the choice among allophones is seldom under conscious control, few people realize their existence. English-speakers may be unaware of the differences among six allophones of the phoneme /t/: unreleased [ t̚] as in cat, aspirated [tʰ] as in top, glottalized [ʔ] as in button, flapped [ɾ] as in American English water, nasalized flapped [ɾ̃] as in winter, and none of the above [t] as in stop. However, they may become aware of the differences if, for example, they contrast the pronunciations of the following words:
- Night rate: unreleased [ˈnʌɪt̚.ɹʷeɪt̚] (without a word space between [ . ] and [ɹ])
- Nitrate: aspirated [ˈnaɪ.tʰɹ̥eɪt̚] or retracted [ˈnaɪ.tʃɹʷeɪt̚]
A flame that is held before the lips while those words are spoken flickers more for the aspirated nitrate than for the unaspirated night rate. The difference can also be felt by holding the hand in front of the lips. For a Mandarin-speaker, for whom /t/ and /tʰ/ are separate phonemes, the English distinction is much more obvious than for an English-speaker, who has learned since childhood to ignore the distinction.
Allophones of English /l/ may be noticed if the 'light' [l] of leaf [ˈliːf] is contrasted with the 'dark' [ɫ] of feel [ˈfiːɫ]. Again, the difference is much more obvious to a Turkish-speaker, for whom /l/ and /ɫ/ are separate phonemes, than to an English speaker, for whom they are allophones of a single phoneme.
These descriptions are more sequentially broken down in the next section.
Rules for English Consonant Allophones
Peter Ladefoged, a renowned phonetician, clearly explains the consonant allophones of English in a precise list of statements to illustrate the language behavior. Some of these rules apply to all the consonants of English; the first item on the list deals with consonant length, items 2 through 18 apply to only selected groups of consonants, and the last item deals with the quality of a consonant. These descriptive rules are as follows:
- Consonants are longer when at the end of a phrase. This can be easily tested by recording a speaker saying a sound like “bib”, then comparing the forward and backward playback of the recording. One will find that the backward playback does not sound like the forward playback because the production of what is expected to be the same sound is not identical.
- Voiceless stops /p,t,k/ are aspirated when they come at the beginning of a syllable, such as in words like "pip, test, kick" [phɪp, thɛst, khɪk]. You can compare this with voiceless stops that are not syllable initial like "stop" [stɑp]. The /t/ voiceless stop follows the /s/ (fricative) here.
- Voiced obstruents, which include stops and fricatives, such as /b,d,g,v,ð,z,ʒ/, that come at the end of an utterance like /v/ in "improve" or before a voiceless sound like /d/ in "add two") are only briefly voiced during the articulation.
- Voiced stops and affricates /b,d,g,dʒ/ in fact occur as voiceless at the beginning of a syllable unless immediately preceded by a voiced sound, in which the voiced sound carries over.
- Approximants (in English, these include /w,r,j,l/) are partially voiceless when they occur after syllable-initial /p,t,k/ like in "play, twin, cue" [pl̥eɪ, tw̥ɪn, kj̥u].
- Voiceless stops /p,t,k/ are not aspirated when following after a syllable initial fricative, such as in the words "spew, stew, skew."
- Voiceless stops and affricates /p,t,k,tʃ/ are longer than their voiced counterparts /b,d,g,dʒ/ when situated at the end of a syllable. Try comparing "cap" to "cab" or "back" to "bag".
- When a stop comes before another stop, the explosion of air only follows after the second stop, illustrated in words like "apt" [æp̚t] and "rubbed" [rʌb̚d].
- Many English accents produce a glottal stop in syllables that end with voiceless stops. Some examples include pronunciations of "tip, pit, kick" [tɪʔp, pɪʔt, kɪʔk].
- Some accents of English use a glottal stop in place of a /t/ when it comes before an alveolar nasal in the same word (as opposed to in the next word), such as in the word "beaten" [ˈbiːʔn̩].
- Nasals become syllabic, or their own syllable, only when immediately following an obstruent (as opposed to just any consonant), such as in the words "leaden, chasm" [ˈlɛdn̩, ˈkæzm̩]. Take in comparison "kiln, film"; in most accents of English, the nasals are not syllabic.
- The lateral /l/, however, is syllabic at the end of the word when immediately following any consonant, like in "paddle, whistle" [ˈpædl̩, ˈwɪsl̩].
- When considering /r,l/ as liquids, /r/ is included in this rule as well as present in the words "sabre, razor, hammer, tailor" [ˈseɪbr̩, ˈreɪzr̩, ˈhæmr̩, ˈteɪlr̩].
- Alveolar stops become voiced taps when they occur between two vowels, as long as the second vowel is unstressed. Take for instance mainly American English pronunciations like "fatty, data, daddy, many" [ˈfæɾi, ˈdæɾə, ˈdæɾi, ˈmɛɾ̃i].
- When an alveolar nasal is followed by a stop, the /t/ is lost and a nasal tap occurs, causing "winter" to sound just like "winner" or "panting" to sound just like "panning". In this case, both alveolar stops and alveolar nasal plus stop sequences become voiced taps after two vowels when the second vowel is unstressed. This can vary among speakers, where the rule does not apply to certain words or when speaking at a slower pace.
- All alveolar consonants assimilate to dentals when occurring before a dental. Take the words "eighth, tenth, wealth". This also applies across word boundaries, for example "at this" [ˈæt̪ ðɪs]].
- Alveolar stops are reduced or omitted when between two consonants. Some examples include "most people" (can be written either as [ˈmoʊs ˈpipl̩] or [ˈmoʊst ˈpipl̩] with the IPA, where the [t] is inaudible, and "sand paper, grand master", where the [d] is inaudible.
- A consonant is shortened when it is before an identical consonant, such as in "big game" or "top post".
- A homorganic voiceless stop may be inserted after a nasal before a voiceless fricative followed by an unstressed vowel in the same word. For example, a bilabial voiceless plosive /p/ can be detected in the word "something" [ˈsʌmpθɪŋ] even though it is orthographically not indicated. This is known as epenthesis. However, the following vowel must be unstressed.
- Velar stops /k,g/ become more front when the following vowel sound in the same syllable becomes more front. Compare for instance "cap" [kæp] vs. "key" [ki] and "gap" [gæp] vs. "geese" gis].
- The lateral /l/ is velarized at the end of a word when it comes after a vowel as well as before a consonant. Compare for example "life" [laɪf] vs. "file" [faɪɫ] or "feeling" [fiːlɪŋ] vs. "feel" [fiːɫ].
There are many examples for allophones in languages other than English. Typically, languages with a small phoneme inventory allow for quite a lot of allophonic variation: examples are Hawaiian and Toki Pona. Here are some examples (the links of language names go to the specific article or subsection on the phenomenon):
- Consonant allophones
- Final devoicing, particularly final-obstruent devoicing: Arapaho, English, Nahuatl, Catalan and many others
- Voicing of initial consonant
- Anticipatory assimilation
- Aspiration changes: Algonquin
- Frication between vowels: Dahalo
- Lenition: Manx, Corsican
- Voicing of clicks: Dahalo
- Allophones for /b/: Arapaho, Xavante
- Allophones for /d/: Xavante
- Allophones for /f/: Bengali
- Allophones for /j/: Xavante
- Allophones for /k/: Manam
- Allophones for /pʰ/: Garhwali
- [g] and [q] as allophones: a number of Arabic dialects
- [l] and [n] as allophones: Some dialects of Hawaiian, and some of Mandarin (e.g. Southwestern and Lower Yangtze)
- Allophones for /n/
- Allophones for /r/: Xavante
- Allophones for /ɽ/: Bengali
- Allophones for /s/: Bengali, Taos
- [t] and [k] as allophones: Hawaiian
- Allophones for /w/:
- Allophones for /z/: Bengali
- Vowel allophones
- [e] and [o] are allophones of /i/ and /u/ in closed final syllables in Malay and [ɪ] and [ʊ] are allophones of /i/ and /u/ in Indonesian.
- [ʉ, ʊ, o̞, o] as allophones for short /u/, and [ɨ, ɪ, e̞, e] as allophones for short /i/ in various Arabic dialects (long /uː/, /oː/, /iː/, /eː/ are separate phonemes in most Arabic dialects).
- Allophones for /i/, /a/ and /u/: Nuxálk
- Vowel/consonant allophones
Representing a phoneme with an allophone
Since phonemes are abstractions of speech sounds, not the sounds themselves, they have no direct phonetic transcription. When they are realized without much allophonic variation, a simple broad transcription is used. However, when there are complementary allophones of a phoneme, the allophony becomes significant and things then become more complicated. Often, if only one of the allophones is simple to transcribe, in the sense of not requiring diacritics, that representation is chosen for the phoneme.
However, there may be several such allophones, or the linguist may prefer greater precision than that allows. In such cases, a common convention is to use the "elsewhere condition" to decide the allophone that stands for the phoneme. The "elsewhere" allophone is the one that remains once the conditions for the others are described by phonological rules.
For example, English has both oral and nasal allophones of its vowels. The pattern is that vowels are nasal only before a nasal consonant in the same syllable; elsewhere, they are oral. Therefore, by the "elsewhere" convention, the oral allophones are considered basic, and nasal vowels in English are considered to be allophones of oral phonemes.
In other cases, an allophone may be chosen to represent its phoneme because it is more common in the languages of the world than the other allophones, because it reflects the historical origin of the phoneme, or because it gives a more balanced look to a chart of the phonemic inventory.
An alternative, which is commonly used for archiphonemes, is to use a capital letter, such as /N/ for [m], [n], [ŋ].
In rare cases, a linguist may represent phonemes with abstract symbols, such as dingbats, to avoid privileging any particular allophone.
- R. Jakobson, Structure of Language and Its Mathematical Aspects: Proceedings of symposia in applied mathematics, AMS Bookstore, 1980, ISBN 978-0-8218-1312-6,
...An allophone is the set of phones contained in the intersection of a maximal set of phonetically similar phones and a primary phonetically related set of phones....
- B.D. Sharma, Linguistics and Phonetics, Anmol Publications Pvt. Ltd., 2005, ISBN 978-81-261-2120-5,
... The ordinary native speaker is, in fact, often unaware of the allophonic variations of his phonemes ...
- Y. Tobin, Phonology as human behavior: theoretical implications and clinical applications, Duke University Press, 1997, ISBN 978-0-8223-1822-4,
...always found that native speakers are clearly aware of the phonemes of their language but are both unaware of and even shocked by the plethora of allophones and the minutiae needed to distinguish between them....
- Lee, Penny (1996). The Whorf Theory Complex — A Critical Reconstruction. John Benjamins. pp. 46, 88.
- Trager, George L. (1959). "The Systematization of the Whorf Hypothesis". Anthropological Linguistics. Operational Models in Synchronic Linguistics: A Symposium Presented at the 1958 Meetings of the American Anthropological Association. 1 (1): 31–35. JSTOR 30022173.
- Hymes, Dell H.; Fought, John G. (1981). American Structuralism. Walter de Gruyter. p. 99.
- Barbara M. Birch, English L2 reading: getting to the bottom, Psychology Press, 2002, ISBN 978-0-8058-3899-2,
...When the occurrence of one allophone is predictable when compared to the other, as in this case, we call this complementary distribution. Complementary distribution means that the allophones are 'distributed' as complements....
- Ladefoged, Peter (2001). A Course in Phonetics (4th ed.). Orlando: Harcourt. ISBN 0-15-507319-2. p. 56-60.
- Hale, Mark (2000). "Marshallese phonology, the phonetics-phonology interface and historical linguistics". The Linguistic Review. 17 (2–4): 241-258.