In human language, a phoneme is a set of phones (speech sounds or sign elements) that are cognitively equivalent. It is the basic unit that distinguishes between different words or morphemes — changing an element of a word from one phoneme to another produces either a different word or obvious nonsense, whereas changing an element from one phone to another, when both belong to the same phoneme, produces the same word (sometimes with an odd or incomprehensible pronunciation).
Phonemes are not the physical segments themselves, but mental abstractions of them. A phoneme could be thought of as a family of related phones, called allophones, that the speakers of a language think of, and hear or see, as being categorically the same.
In sign languages, the phoneme was formerly called a chereme (or cheireme), but usage changed to phoneme when it was recognized that the mental abstractions involved are essentially the same as in oral languages.
Phonemics, a branch of phonology, is the study of the systems of phonemes of languages.
Although the concept has been fundamental to the development of phonological analysis of language beneath the level of the syllable, some linguists reject the theoretical validity of the phoneme. Some think that phonemes are more a product of literacy (i.e., the need to categorize the phonetics of a language in order to write it down systematically with a minimum number of letters). Other critics charge that the mind processes sub-phonemic elements of speech (e.g., features) in meaningful ways.
A common test to determine whether two phones are allophones or separate phonemes relies on finding so-called minimal pairs: words that differ only in the phones in question.
The term phonème was reportedly first used by Dufriche-Desgenettes in 1873, but it refered to only a sound of speech. The term phoneme as an abstraction was developed by the Polish linguist Jan Niecislaw Baudouin de Courtenay and his student Mikołaj Kruszewski during 1875-1895. The term used by these two was fonema, the basic unit of what they called psychophonetics. The concept of the phoneme was elaborated in the works of Nikolai Trubetzkoi and other of the Prague School (during the years 1926-1935), as well as in that of structuralists like Ferdinand de Saussure, Edward Sapir, and Leonard Bloomfield. Later, it was also used in generative linguistics, most famously by Noam Chomsky and Morris Halle, and remains central in any accounts of the development of virtually all modern schools of phonology.
The phoneme can be defined as "the smallest meaningful psychological unit of sound." The phoneme has mental, physiological, and physical substance: our brains process the sounds; the sounds are produced by the human speech organs; and the sounds are physical entities that can be recorded and measured.
For an example of phonemes, consider the English words pat and sat, which appear to differ only in their initial consonants. This difference, known as contrastiveness or opposition, is sufficient to distinguish these words, and therefore the P and S sounds are said to be different phonemes in English. A pair of words that are identical except for such a sound are known as a minimal pair; this is the most frequent demonstration that two sounds are separate phonemes.
If no minimal pair can be found to demonstrate that two sounds are distinct, it may be that they are allophones. Allophones are variant phones (i.e., sounds) that are not recognized as distinct by a speaker, and are not meaningfully different in the language, and so are perceived as "the same". This is especially likely if they consistently occur in different environments. For example, the "dark" L sound at the end of the English word "wool" is quite different from the "light" L sound at the beginning of the word "leaf", but this difference is meaningless in English, and is determined by whether the sound is at the beginning or end of a word. A native English speaker might have a hard time hearing the difference at first, but in Turkish the difference between "light" and "dark" L is sufficient to distinguish words. That is, they are two separate phonemes in Turkish, but allophones of a single phoneme in English.
The phonemic relationship of two sounds may not be obvious to a non-native speaker, which is why minimal pairs and an understanding of phonetic environments are important. For example, in Korean, there is a phoneme /r/ that is a flapped r between vowels, and is an l-sound next to other consonants. These sound very different to an English speaker, who is attuned to hearing them because the differences are meaningful in English. However, the native speaker has learned from an early age to filter out the difference, as they are not meaningful in their language. In Korean, for instance, it is impossible to distinguish the two words "ram" and "lam", despite the fact that both R and L sounds occur in the language.
Across multiple languages, the same IPA symbol may be used to represent a phoneme, but their actual pronunciation may not be identical but merely similar. For example, the Finnish word maat ("countries") sounds different from the British English (Received Pronunciation) word mart even though both are phonemically transcribed as IPA /mɑ:t/; the Spanish word sin ("without") has a somewhat different vowel from the American English seen though both are transcribed as /sin/. Such distinctions can be made in a phonetic transcription.
The exact number of phonemes in English depends on the speaker and the method of determining phoneme vs. allophone, but estimates typically range from 40 to 45, which is above average across all languages. Pirahã has only 10, while !Xóõ has 141.
Some languages make use of pitch for the precise same purpose. In this case, the tones used are called tonemes. Some languages distinguish words made up of the same phonemes (and tonemes) by using different durations of some elements, which are called chronemes. However, not all scholars working on languages with distinctive duration use this term.
In sign languages, phonemes may be classified as Tab (elements of location, from Latin tabula), Dez (the hand shape, from designator), Sig (the motion, from signation), and with some researchers, Ori (orientation). Facial expressions and mouthing are also phonemic.
A transcription that only indicates the different phonemes of a languages is said to be phonemic. Such transcriptions are enclosed within virgules (slashes), / /; these show that each enclosed symbol is claimed to be phonemically meaningful. On the other hand, a transcription that indicates finer detail, including allophonic variation like the two English L's, is said to be phonetic, and is enclosed in square brackets, [ ].
The common notation used in linguistics employs virgules (slashes) (/ /) around the symbol that stands for the phoneme. For example, the phoneme for the initial consonant sound in the word "phoneme" would be written as /f/. In other words, the graphemes are <ph>, but this digraph represents one sound /f/. Allophones, more phonetically specific descriptions of how a given phoneme might be commonly instantiated, are often denoted in linguistics by the use of diacritical or other marks added to the phoneme symbols and then placed in square brackets ([ ]) to differentiate them from the phoneme in slant brackets (/ /). The conventions of orthography are then kept separate from both phonemes and allophones by the use of the markers < > to enclose the spelling.
The symbols of the International Phonetic Alphabet (IPA) and extended sets adapted to a particular language are often used by linguists to write phonemes of oral languages, with the principle being one symbol equals one categorical sound. Due to problems displaying some symbols in the early days of the Internet, systems such as X-SAMPA and Kirshenbaum were developed to represent IPA symbols in plain text. As of 2004, any modern web browser can display IPA symbols (as long as the operating system provides the appropriate fonts), and we use this system in this article.
The only published set of phonemic symbols for a sign language is the Stokoe notation developed for American Sign Language, which has since been applied to British Sign Language by Kyle and Woll, and to Australian Aboriginal sign languages by Adam Kendon. However, there are several phonetic systems, such as SignWriting.
Examples of phonemes in the English language would include sounds from the set of English consonants, like /p/ and /b/. These two are most often written consistently with one letter for each sound. However, phonemes might not be so apparent in written English, such as when they are typically represented with combined letters, called digraphs, like <sh> (pronounced /ʃ/) or <ch> (pronounced /tʃ/).
To see a list of the phonemes in the English language, see IPA for English.
Two sounds that may be allophones (sound variants belonging to the same phoneme) in one language may belong to separate phonemes in another language or dialect. In English, for example, /p/ has aspirated and non-aspirated allophones:aspirated as in /pɪn/, and non-aspirated as in /spɪn/. However, in many languages (e. g. Chinese), aspirated /pʰ/ is a phoneme distinct from unaspirated /p/. As another example, there is no distinction between [r] and [l] in Japanese, there is only one /r/ phoneme in Japanese, although the Japanese /r/ has allophones that make it sound more like an /l/, /d/, or /r/ to English speakers. The sounds /z/ and /s/ are distinct phonemes in English, but allophones in Spanish. /n/ (as in run) and /ŋ/ (as in rung) are phonemes in English, but allophones in Italian and Spanish.
An important phoneme is the chroneme, a phonemically-relevant extension of the duration of a consonant or vowel. Some languages or dialects such as Finnish or Japanese allow chronemes after both consonants and vowels. Others, like Italian or Australian English use it after only one (in the case of Italian, consonants; in the case of Australian, vowels).
Arguments against the phoneme
Rather than a basic mental unit of language, some think that the phoneme may well be a perceptual artifact of alphabetic literacy (see the terms Phonemic awareness and Phonological awareness). If not that, it may be an epiphenomenal aspect to listening removed from face-to-face encounters, that is, text-like listening (qv phone and feature). It could be said that the unit of the phoneme is a necessary construct if we wish to set a dynamic, complex spoken language into static, written form expressed at a sub-syllabic level, though the model is a simplification and no where near phonologically or phonetically complete. The phoneme has the theoretical weakness from the perspective of phonology in that it uses, in part, lexical criteria to determine something that is supposed to be phonological (i.e., minimal pairs of words to point out phonological categories).
Much of phonology, while accepting the phoneme as possible model or unit of language for description, has largely moved past the segmental phoneme as a basic unit of speech, of speech processing or of language acquisition. This is because the concept of the 'feature' is viewed as beneath the level of the phoneme while also spanning across segments. Meanwhile, attempts at capturing a phonological picture of the psychological control and structure underlying real speech flounder on the inadequacies of the phoneme for such purposes (that is, the phoneme can not account for co-articulation or assimilation of controlled speech, among other phenomena). Such an endeavor is more for the field of articulatory phonology, and its rival unit of phonology is the 'articulatory gesture'. However, the term 'phoneme', though variably defined and delimited, remains a widely and uncritically accepted concept in second and foreign language teaching and in the psychology of native literacy (especially for acquisitional literacy in alphabetic languages, such as English).
A restricted phoneme is a phoneme that can only occur in a certain environment: There are restrictions as to where it can occur. English has several restricted phonemes:
- /ŋ/, as in sing, occurs only at the end of a syllable, never at the beginning. (In many other languages, such as Swahili, /ŋ/ can start a word.)
- /h/ occurs only at the beginning of a syllable, never at the end. (A few languages such as Arabic allow /h/ at the ends of words.)
- In many American dialects with the cot-caught merger, /ɔ/ occurs only before /r/, /l/, and in the diphthong /ɔi/.
- In non-rhotic dialects, /r/ can only occur before a vowel, never at the end of a word or before a consonant.
- Under most interpretations, /w/ and /j/ occur only before a vowel, never at the end of a syllable. However, many phonologists interpret a word like boy as either [boi̯] or [boj].
Neutralization, archiphoneme, underspecification
Phonemes that are contrastive in certain environments may not be contrastive in all environments. In the environments where they don't contrast, the contrast is said to be neutralized.
An example from English is the neutralization of the plosives /k, g/ following /s/. Phonetically, the unaspirated tenuis plosive in sky is closer to English /g/, which is partially voiceless in the initial position, than to aspirated /k/. This can be heard by comparing the sky with this guy; also, in the speech of young children who are not yet able to produce consonant clusters, they often pronounce sky as what sounds like /gai/ to adult ears but is actually produced with an unaspirated, unvoiced [k]. That is, /k/ and /g/ are constrastive word initially,
But not after an /s/,
Thus one cannot say whether the underlying representation of the plosive in sky is /k/ without aspiration, or /g/ without voicing. This neutralization can instead be represented as an archiphoneme |G|, in which case the underlying representation of sky would be |sGaɪ|.
Another way to talk about archiphonemes involves the concept of underspecification. Phonemes can be considered fully specified segments while archiphonemes are underspecified segments. In Tuvan, phonemic vowels are specified with the features of tongue height, backness, and lip rounding. The archiphoneme |U| is an underspecified high vowel where only the tongue height is specified.
Whether |U| is pronounced as front or back and whether rounded or unrounded depends on vowel harmony. If |U| occurs following a front unrounded vowel, it will be pronounced as the phoneme /i/; if following a back unrounded vowel, it will be as an /ɯ/; and if following a back rounded vowel, it will be an /u/. This can been seen in the following words:
|-|Um|||'my'||(the vowel of this suffix is underspecified)|
||idikUm|||→||[idikim]||'my boot'||(/i/ is front & unrounded)|
||xarUm|||→||[xarɯm]||'my snow'||(/a/ is back & unrounded)|
||nomUm|||→||[nomum]||'my book'||(/o/ is back & rounded)|
It should be noted that not all phonologists accept the concept of archiphonemes. Many doubt that it reflects how people process language or control speech, and some argue that archiphonemes add unnecessary complexity.
Prothesis, epenthesis and paragoge, due to phonotactics, add sounds into words without adding meaning. Nevertheless, the sound is added, and thus the phoneme status may be ambiguous. For example, in Spanish a prothetic e- must be added before initial /s/ + consonant clusters, e.g. estrés.
Of all the sounds that a human vocal tract can create, different languages vary considerably in the number of these sounds that are considered to be distinctive phonemes in the speech of that language. Ubyx and some dialects of Abkhaz Abkhaz have only two phonemic vowels, and many Native American languages have three. On other extreme, the Bantu language Ngwe has fourteen vowel qualities, twelve of which may occur long or short, for twenty-six oral vowels, plus six nasalized vowels, long and short, for thirty-eight vowels; while !Xóõ achieves thirty-one pure vowels—not counting vowel length, which it also has—by varying the phonation. Rotokas has only six consonants, while !Xóõ has somewhere in the neighborhood of seventy-seven, and Ubyx eighty-one. French has no phonemic tone or stress, while several of the Kam-Sui languages have nine tones, and one of the Kru languages, Wobe, has been claimed to have fourteen, though this is disputed. The total number of phonemes in languages varies from as few as eleven in Rotokas to as many as 112 in !Xóõ (including four tones). These may range from familiar sounds like [t], [s], or [m] to very unusual ones produced in extraordinary ways (see: Click consonant, phonation, airstream mechanism). The English language itself uses a rather large set of thirteen to twenty-two vowels, including diphthongs, though its twenty-two to twenty-six consonants are close to average. (There are twenty-one consonant and five vowel letters in the English alphabet, but this does not correspond to the number of consonant and vowel sounds.)
The most common vowel system consists of the five vowels /i/, /e/, /a/, /o/, /u/. The most common consonants are /p/, /t/, /k/, /m/, /n/. A very few languages lack one of these: standard Hawai‘ian lacks /t/, Mohawk lacks /p/ and /m/, Hupa lacks both /p/ and a simple /k/, colloquial Samoan lacks /t/ and /n/, while Rotokas and Quileute lack /m/ and /n/. While most of these languages have very small inventories, Quileute and Hupa have quite complex consonant systems.
At least in theory, in a phonemic writing system, a given symbol represents a single phoneme, and each phoneme is represented by a single symbol. This may differ from a phonetic orthography, which only requires that the spelling be unambiguously determined by the pronunciation, and the pronunciation unambiguously indicated by the spelling. Phonemic representation of a language is often described as 'broad transcription', while a phonetic rendering is called 'narrow'. A phonetic system would have more symbols or spelling conventions, since it might, in part, attempt to capture some key sound variations (allophones)of a phoneme. Learners of a foreign or second language can benefit from a more phonetic writing system if it reveals subtleties in pronunciation that are phonemically glossed over by literate native or fluent speakers of that language (since the latter's purpose is fluent reading).
English spelling (whether British, American or Australian) is often cited as the classic example of a nonphonemic, and indeed unphonetic, spelling system. Welsh and Irish are, by contrast, among the more predictable orthographies among languages using the Latin alphabet. In French, rules to predict pronunciation from spelling are quite simple and have few exceptions, as long as there are some clues such as context or part of speech, but guessing spelling from pronunciation is quite difficult, especially because of the many silent letters. It should be noted that both written English and French (being lexical cousins, if quite different phonologically speaking) tend to preserve word root (over sound) relationships. Italian, Spanish and especially Finnish have a very close letter-to-phoneme correspondence. Karelian has a perfectly phonemic spelling system, as it has no standard language, but it has a complete spelling system.
Other languages fall somewhere in between polar distinctions such as "lexical vs. phonemic and/or phonetic" and "phonemic vs. phonetic". Although English is often given as an example of an unphonetic orthography, its system is nowhere near to being as logographic (lexical or word-based) a system as Chinese writing is. English spelling conveys etymological, derivational and inflectional information, but also vast amounts of phonetic information as well. In a nutshell, written English displays a great deal of complexity for representing vowel sounds within a fairly stable and consistent consonant framework (though there is a shortage of letters all around, with 26 letters and phoneme counts well over 40). Spanish is often given as an example of a phonetic orthography, but it has numerous imperfections including silent letters. It is, at least, possible to tell the correct pronunciation of any written Spanish word. Another phonemic orthography is Serbian. Its phonemicity was established by Serbian "Webster" Vuk Stefanović Karadžić. He followed a strict phonemic principle, which is best told by his own words: "Write as you speak and read as it is written.". Hindi, a descendant of Sanskrit, is an example of phonetically written language represented with a non-Roman Alphabet that is partly syllabic in nature. Hindi's writing system, however, probably ultimately descends from the same ancient Middle Eastern sources that gave the world the Roman, Cyrillic and Arabic scripts.
Real world distinctions between phonemic and nonphonemic orthographies are exaggerated. All languages are written with conventions that represent both meaning and pronunciation. This is true at both ends of the scale: Chinese characters are first and foremost symbols for morphemes and words, but they may have some phonetic elements to their composition as well (and these work, sometimes at least, the way spelling analogies do in written English). At the other extreme, there are a few orthographies which are complete and consistent phonemic representations of an artificial national standard. The phonemic principles by which orthographies might be standardized might also exclude representation of variations in pronunciation within the spoken dialects of a national language.
Korean hangul, an invented writing system, has been called a linguistically perfect writing system because of its attention to phonetic-featural detail, capturing the language analytically at a fine-tuned featural level. However, it is not clear that native speakers need such phonetic-featural detail to learn to read Korean, and Korean, as it is actually written and read, could be processed synthetically at higher levels language, such as syllable types and whole words (i.e., sight words that are quickly read because they occur so often in text).
- What is a phoneme? (SIL)
- What is an allophone? (SIL)
- What is a phone? (SIL)
- What is a phonetically similar segment? (SIL)
- What is a minimal pair? (SIL)
- What is complementary distribution? (SIL)
- What is an environment? (SIL)
- What is an contrast in identical environments? (SIL)
- What is an contrast in analogous environments? (SIL)
- Comparison of morpheme-morph-allomorph & phoneme-phone-allophone? (SIL)
- What is phonology? (SIL)
- Phoneme (Lexicon of Linguistics)
- Allophony (Lexicon of Linguistics)
- Transcription (Lexicon of Linguistics)
- Grapheme-Phoneme Conversion (Lexicon of Linguistics)
- Phoneme Restoration (Lexicon of Linguistics)
- phonemic awareness