Phoneme

• 1. Introduction

  ›

  A phoneme is the smallest unit of sound a language treats as meaningfully distinct, the difference between the word cat and the word bat. Some languages get by with astonishingly few of these units: Pirahã and Rotokas each function with somewhere around nine to eleven phonemes in their entire sound system. The language ǃXũ, by contrast, uses as many as 141. Both extremes describe fully functional human languages, capable of saying anything their speakers need to say. What is a phoneme, exactly, and how can two languages differ so wildly in how many building blocks they use to construct every word they will ever speak?

• 2. Building Blocks Of Words

  ›

  The English words cell and set share every sound except their final consonant, written in the International Phonetic Alphabet as /sɛl/ versus /sɛt/, and because /l/ and /t/ alone separate those two words, linguists count them as phonemes of English. Spelling does not always track phonemes closely: the words knot, nut, and gnat all share the consonant phonemes /n/ and /t/, differing only in their vowels, /ɒ/, /ʌ/, and /æ/. A longer example, the spoken word pushed, breaks down into a sequence of four phonemes, /p/, /ʊ/, /ʃ/, and /t/. Linguists write phonemes between forward slashes, as in /k/ for the sound at the start of cat, and reserve square brackets for a more precise phonetic transcription that captures fine detail, including variation that does not change meaning at all. The Latin alphabet was devised specifically for Classical Latin, which gave that language's writing system a near one-to-one match between phonemes and letters, though even its inventors chose not to mark vowel length. English uses a two-letter digraph, sh, to spell the single phoneme /ʃ/, and conversely lets a single letter, x, stand for the two-phoneme sequence /gz/ or /ks/, a reminder that spelling and phonemes rarely line up perfectly.

• 3. One Sound, Many Shades

  ›

  English treats the sounds written and as distinct phonemes because they separate sin from sing, /sɪn/ from /sɪŋ/, but Spanish speakers hear no such distinction, treating pan and paŋ as regional pronunciations of the same word, pan, meaning bread. Such variant pronunciations of a single phoneme are called allophones. The English p in spit is unaspirated, written p, while the p in pit carries an extra burst of air and is written pʰ, yet both are heard by English speakers as the same phoneme, /p/. American English shows the same flexibility with the letter t: it is a glottal stop in cat, an alveolar flap in dating, a plain plosive in stick, and an aspirated plosive in tie, four distinct sounds that speakers nonetheless perceive as a single phoneme, /t/. Icelandic draws the line differently. There, kʰ begins the word for cheerful and k begins the word for riddles, so substituting one sound for the other changes the meaning, meaning Icelandic treats /kʰ/ and /k/ as two separate phonemes rather than one. Languages also restrict where a given phoneme is allowed to occur: in English, /ŋ/, the sound at the end of sing, appears only at the end of a syllable and never at the start, even though languages such as Māori, Swahili, Tagalog, Thai, and Setswana permit it word-initially. English treats /h/ the opposite way, allowing it only at the start of a syllable, while a few languages, including Arabic and Romanian, allow /h/ at the end of one instead.

• 4. Testing With Minimal Pairs

  ›

  The Icelandic words kátur and gátur differ in only one sound, kʰ against k, making them what linguists call a minimal pair, and the existence of such pairs is the standard test for whether two sounds belong to different phonemes or are merely allophones of the same one. English offers its own case in tip and dip, where the single sound difference between t and d changes the meaning entirely, confirming /t/ and /d/ as separate phonemes. American Sign Language has minimal pairs too, differing in just one of five parameters: handshape, movement, location, palm orientation, or a nonmanual marker. Sometimes no true minimal pair exists and linguists rely on a near minimal pair instead; English has no exact minimal pair distinguishing ʃ from ʒ, but the words pressure and pleasure come close enough to demonstrate the two are separate phonemes, while less common words such as Confucian and confusion supply an exact minimal pair for the same contrast.

• 5. Stress And Tone

  ›

  English contains two words spelled invite, one a verb stressed on its second syllable, /ɪnˈvaɪt/, the other a noun stressed on its first, /ˈɪnvaɪt/, with the position of stress alone marking the difference in meaning. French cannot use stress this way, because word stress in French is generally predictable from position and therefore carries no meaning of its own, which is why dictionaries do not bother marking it. Mandarin Chinese instead relies on tone: the syllable ma can mean mother, hemp, horse, scold, or serve as a question particle, depending entirely on which of five tonal pronunciations a speaker uses. Languages that use tone this way are said to have phonemic tone, and their tone contrasts are sometimes called tonemes, while English, lacking phonemic tone, uses intonation instead for functions such as emphasis.

• 6. A New Word For An Old Idea

  ›

  The word phonème was reportedly first used by A. Dufriche-Desgenettes in 1873, though at that point it referred only to a speech sound rather than an abstract category. The idea of the phoneme as an abstraction was developed between 1875 and 1895 by the Polish linguist Jan Baudouin de Courtenay and his student Mikołaj Kruszewski, who called their basic unit fonema, part of what they termed psychophonetics. Daniel Jones became the first linguist in the western world to use the word phoneme in its modern sense, in his article The Phonetic Structure of the Sechuana Language. The concept was then elaborated between 1926 and 1935 by Nikolai Trubetzkoy and the Prague School, and by structuralists including Ferdinand de Saussure, Edward Sapir, and Leonard Bloomfield, before Noam Chomsky and Morris Halle later redefined it within generative linguistics, keeping the phoneme central to modern phonology even as newer models supplemented it. Chomsky and Halle also challenged an older structuralist requirement called biuniqueness, the rule that a given phone must always map to one and only one phoneme; the flapped sound heard in the middle of hitting and bidding realizes /t/ in the first word and /d/ in the second, a case that seemed to break that rule and fueled debate through the late 1950s and early 1960s. Linguists have also disagreed about whether a language even has one correct phonemic analysis at all: Kenneth Pike argued there is only one accurate phonemic analysis for a given set of data, while Yuen Ren Chao, writing in 1934, countered that the sounds of a language can usually be reduced to phonemes in more than one valid way, some better suited to certain purposes than others. The linguist F. W. Householder later labeled that split God's Truth against hocus-pocus, and the disagreement was never fully settled.

• 7. Signs As Sound

  ›

  William Stokoe, working at Gallaudet University, coined the terms cherology and chereme in 1960 while arguing that sign languages were true and full languages in their own right, a position once controversial and now universally accepted in linguistics. Stokoe identified three bundles of articulation making up a sign: tab, the location, from the Latin tabula; dez, the handshape, from designator; and sig, the motion, from signation, with later researchers adding orientation and facial expression to the model. In American Sign Language, the signs for father and mother differ minimally in location while handshape and movement stay identical, making location the contrastive element, exactly as a single consonant can distinguish two spoken words. Stokoe's own terminology has since been largely abandoned in favor of simply calling these units phonemes, a shift that underscores how functionally and psychologically equivalent signed and spoken languages turned out to be.

• 8. Counting The World's Sounds

  ›

  Pirahã functions with as few as nine to eleven phonemes total and Rotokas with eleven, while ǃXũ uses as many as 141, illustrating just how far the world's languages spread across the possible range. Vowel counts vary just as widely: Ubykh and Arrernte get by with only two distinct vowels, while the Bantu language Ngwe has 14 vowel qualities, twelve of which can be long or short, producing 26 oral vowels plus six nasalized ones for a total of 38, and !Xóõ achieves 31 pure vowels before counting additional variation in length. Consonant inventories show similar extremes: Puinave and the Papuan language Tauade each use just seven consonants, Rotokas gets by with six, while !Xóõ has around 77 and Ubykh has 81. English sits above average on vowels, with 13 to 21 vowel phonemes including diphthongs, but close to average on consonants, with 22 to 26; broken down by accent, one standard analysis counts 20 vowel phonemes in Received Pronunciation, 14 to 16 in General American, and 20 to 21 in Australian English, though other analyses of the very same recordings reach different totals entirely. Across all languages, the average works out to about 22 consonant phonemes and 8 vowel phonemes, and the five vowels /i/, /e/, /a/, /o/, /u/ and the five consonants /p/, /t/, /k/, /m/, /n/ are the most common building blocks of all, though even these are not universal: Arabic lacks /p/, standard Hawaiian lacks /t/, Mohawk and Tlingit lack both /p/ and /m/, and colloquial Samoan gets by without /t/ or /n/ at all.