[Journal of the Simplified Spelling Society, J27, 2000/1 p11-13]
[Note: this page is set to display International Phonetic Alphabet characters. If your browser over-rides this setting, by default or by choice, the IPA characters may not show.]
[See IPA table and link to full IPA chart.]
[See other journal articles by Adam Brown and about fonemes.]

The number of phonemes in English: not a simple answer to a simple question.

Adam Brown.

Dr Brown is Associate Professor at the National Institute of Education, Singapore. For the past 20 years he has been concerned with how the pronunciation of English can best be taught to non-native speakers, and his main publication on the subject (Pronunciation Models, Singapore University Press, 1991) was reviewed in JSSS 1993/1, pp25-26.

0. Abstract.

The alphabetical principle states that the letters in the spelling should represent the phonemes in the pronunciation. This presupposes that the number of phonemes is agreed. This article highlights various factors which make this a problematic procedure for English: historical change, sociolinguistic variation, and differences in analysis.

1. Introduction.

The alphabetic principle which underlies the spelling system of English, states that the letters in the spelling should represent the sounds (phonemes) of the pronunciation. Ideally, the correspondence would be one-to-one, as in phonemic transcription. However, while the correspondence in some languages like Finnish and Malay is close to one-to-one, in others (and English is perhaps the worst culprit) it is many-to-one and one-to-many. Nevertheless, that is the underlying principle.

This then begs the question of whether we can establish how many phonemes there are in English. Bett (1999) notes that "in her letter to The Express, Masha Bell said 'We have 256 ways of representing the 45 basic sounds of our language.'" In Allan Campbell's letter to The Press (Christchurch, NZ), he said "There are 41 sounds in English..." In PV7, Steve Bett said "there are 41 significant speech sounds or phonemes." Godfrey Dewey (1970) listed examples of 561 ways that 41 English sounds could be spelled. Bett himself (1999) concluded that "the minimum number of pure phonemes required to accurately transcribe English speech is 34 (12 vowels + 22 consonants)."

The purpose of this article is to elucidate various factors leading to these differences in opinion, and to show that one cannot give a global figure for the number of phonemes in English. These factors are of three types: historical change, sociolinguistic variation, and differences in analysis.

2. Historical change.

It is an axiom of linguistics that languages change over time. This is as true of phonology (including the number of phonemes) as of grammar and vocabulary. Two examples of ongoing changes in British pronunciation will be enough to make the point.

The first concerns words containing the diphthong /ʋə/, which are increasingly changing to /ɔ:/. Wells (1982 : 237) labels this the CURE-FORCE merger, "whereby the /ʋə/ of CURE undergoes a lowering, sometimes via intermediate stages such as [oə] and [ɔə], to [ɔ:], which is identical phonetically with the /ɔ:/ of FORCE, NORTH, THOUGHT. Thus sure comes to be a homophone of shore." Wells (1990 : 547) states that of a panel of 275 British speakers, 57% preferred /pɔ:/ for poor, and 43% /pʋə/, whereas for other words including curious, gourd, tourist, the /ʋə/ pronunciation is given in blue print as recommended for learners of English.

The second example concerns pairs of words such as bored and board. In some British accents, there is a difference, the first being pronounced [bɔəd] and the second [bɔ:d]. This difference has been described by many writers. Jones (1956) treated /ɔə/ as a separate centring diphthong phoneme alongside /ɩə,eə, ʋə/. Wells (1982 : 311) writes that "Minimal pairs such as bored-board may be considered diagnostic for a 'modified London' accent as against non-regional RP [Received Pronunciation]... Accordingly I have treated the opposition as established in London English and I have recognized /ɔə/ and /ɔ:/ as distinct phonemes." The distinction between /ɔə, ɔ:/ seems to have died out in RP, both having become /ɔ:/.

3. Sociolinguistic variation.

Pronunciation varies across geographical accents, perhaps more so for English than for any other language. One major division in English accents is between those (including General American: GA) that pronounce syllable-final /r/ as in car park, and those (including RP) that don't (rhotic vs. non-rhotic). As a result, words such as near, square and cure hold implications for the number of phonemes. In rhotic accents, these words contain a final /r/ preceded by a vowel which is identified as the same phoneme as occurs before other consonants. Thus, kit, dress, foot contain the vowel phonemes /ɩ, e, ʋ/ followed by the consonants /t, s/, while near, square, cure contain the same /ɩ, e, ʋ/ followed by the consonant /r/. In non-rhotic accents, on the other hand, there is no final /r/, and three new vowel phonemes have to be posited: the centring diphthongs /ɩə, eə, ʋə/.

Differences between accents may also take the form of the differentiation in one accent of sounds (especially vowels) which are not differentiated in another (known as differences of phonemic or phonological system: Wells 1982 : 76). A good example of this is the Scottish English pronunciation of low and back vowels. Most Scots do not distinguish pull and pool, and many also conflate Sam and psalm, and cot and caught (Abercrombie 1979). However, these vowels are distinct phonemes in most other accents of English. These Scots therefore have three fewer phonemes here than other accents do.

A GA example is the pronunciation of the words bomb and balm. These are identical (homophones) for GA, whereas in other accents they are distinct, eg, RP /bɒm,bɑ:m/. Other accents therefore have one more phoneme in this area than GA does.

From the above three examples, readers should not jump to the conclusion that RP has more vowel phonemes than other accents of English, and that these other accents are merely simplified versions of RP. Indeed, there are accents which have more phonemes than RP in certain areas. For instance, in the accent of East Anglia (northeast of London), pairs such as moan/mown, sole/soul, nose/knows and toe/tow are not pronounced as homophones, as they are in other accents. Instead, they constitute minimal pairs, the contrast being that the first member of each pair is pronounced with a [ʋu] vowel, while the second has [ʌu] (Wells 1982 : 337). East Anglian English thus has one more phoneme than RP, etc, in this respect.

4. Differences in analysis.

Even where the pronunciation being investigated is a single accent at a single point in time, there may be differences of opinion as to the number of phonemes, owing to differences in analysis. These differences are often of the British-school vs. American-school type (Ladefoged 1993 : 75). An example of this is (what in British school are considered) long vowels and diphthongs, as in bee, boo, bay, buy, boy, (violin) bow, bough. In British school, these are analysed as long monophthong vowel phonemes, or diphthong vowel phonemes, thus /bɨ:, bu:, beɩ, baɩ, bɔɩ, bəʋ, baʋ/. However, in American school (eg, Prator & Robinett 1985), they are usually considered sequences of a vowel phoneme followed by the same consonants that appear at the beginning of yet and wet (for which the symbols /y, w/ are used), thus /biy, buw, bey, bay, bɔy, bow, baw/. In such an analysis, there are no long vowel or diphthong phonemes, and the inventory of vowel phonemes is therefore much smaller.

Another problem in analysis relates to final unstressed vowels. The solution adopted by many reference books and dictionaries nowadays is a compromise which contravenes phonemic theory.
How should we transcribe the words easy and busy as pronounced in RP? ... The possibilities, using our phoneme symbols, are the following: [i:zi:] or [i:zɩ], [bɩzi:] or [bɩzɩ]. Few speakers of RP seem to feel satisfied with any of these transcriptions. There is a possible solution to this problem, but it goes against standard phoneme theory. We can symbolise this weak vowel as [i] that is, using the symbol for the vowel in beat but without the length mark, thus [i:zi, bɩzi]. The [i] vowel is neither the /i:/ of beat nor the /ɩ/ of bit, and is not in contrast with them. We can set up a corresponding vowel [u] for words like value, or unstressed to that is neither the /u:/ of shoe nor the /ʋ/ of book but a weak vowel that shares the characteristics of both. If we use [i] and [u] in our transcription as well as /i:, ɩ, u:, ʋ/, it is no longer a true phonemic transcription in the traditional sense. However, this need not be too serious an objection, and the fact that native speakers seem to think that this transcription fits better with their feelings about the language is a good argument in its favour.
(Roach 1991 : 77-8)
The vowels of words like fire and tower (in non-rhotic accents) may also be analysed differently. As triphthongs [aɩə, aʋə], they may be judged to be one syllable (and thus one phoneme /aɩə, aʋə/) or two syllables (and thus two phonemes /aɩ, aʋ/ + /ə/). Perceptions of this may be affected by the fact that these vowels often undergo a process known as smoothing (Wells 1982 : pp238-242), resulting in the monosyllabic pronunciation [faə, taə] or [fa:, ta:].

A final, and consonantal, difference in analysis relates to the velar nasal [ŋ]. This is undoubtedly a single sound, and the existence of minimal pairs such as sing vs. sin allow the taxonomic phonemic view of phonology to establish a /ŋ/ phoneme. However, some analysts in generative phonology, in particular Chomsky & Halle (1968), have given arguments for positing that surface occurrences of [ŋ] are derived by rule from an underlying sequence /n/ + /g/; in other words, there is no /ŋ/ phoneme - it is only a surface manifestation of different underlying segments.


All of the above instances are situations leading to differences of opinion as to the number of phonemes in English. In short, one cannot give a simple answer to the question. Three other points ought to be made before we close.

Firstly, problems such as the above have always been problems in the taxonomic approach to phonemic theory, partly because the phoneme was not originally established as a theoretical construct. Although several analysts - including such eminent linguists as Isaac Pitman, Edward Sapir, Henry Sweet, Ferdinand de Saussure and Baudouin de Courtenay - used the phoneme as a concept and as a term before him (Abercrombie 1991a), it was Daniel Jones who made the phoneme more widely known. However
Jones always said there was no such thing as phonology as a subject separate from phonetics (he never used the word phonemics). His phoneme concept was unpretentious and unadventurous. Its purpose was to be of service to applied phonetics, especially in the making of transcriptions for language teaching. As Jones wrote in 1931: 'The main object of grouping the sounds of a language into phonemes is to establish a simple and adequate way of writing the language'. Nothing more ambitious was expected of the concept.
(Abercrombie 1991b : 45)
Secondly, while an ideal spelling system for English could be expected to create a one-to-one correspondence between letters and the phonemes of a standard native accent such as GA or RP, it is worth remembering that the majority of speakers of English worldwide nowadays are not native speakers. Moreover, the phonologies of many of these non-native speakers represent simplified systems (ie, contain fewer phonemes) than GA or RP. An ideal spelling system for native speakers may therefore contain significant redundancy for such speakers. Their phonologies represent another factor in establishing the number of phonemes in English.

Thirdly, we may question whether the alphabetic principle (ie, a strict one-to-one correspondence between graphemes and phonemes) is a realistic goal or rather an unattainable ideal for the English spelling system in its current state. There is no denying that certain common features of reformed systems (eg, the elimination of redundant letters) is a move towards a one-to-one correspondence. However, there are other features (eg, the treatment of magic E and of doubled consonant letters) which are attempts primarily to regularize spelling rather than to achieve a one-to-one correspondence. There are clearly other factors than a one-to-one correspondence that need to be taken into account in a reformed system. For instance, /θ, ð/ are undoubtedly separate phonemes, but many reformed systems do not feel obliged to differentiate them in spelling, since they carry a low functional load. A one-to-one correspondence (and its logical prerequisite, knowledge of the number of phonemes) therefore need not necessarily be the sole driving force of spelling reform.


Abercrombie D (1979) 'The accents of standard English in Scotland' in A J Aitken & T McArthur Languages of Scotland Chambers, Edinburgh, pp68-84.

- (1991a) 'Phoneme, the concept and the word' in D Abercrombie Fifty Years in Phonetics Edinburgh University Press, pp22-26.

- (1991b) 'Daniel Jones's teaching' in D Abercrombie Fifty Years in Phonetics Edinburgh University Press, pp37-47.

Bett S (1999) 'Can we pin down the number of phonemes in English?' Simpl Speling (March), p7.

Chomsky N & Halle M (1968) The Sound Pattern of English New York: Harper & Row.

Dewey, G (1970) Relative frequency of English spellings New York: Teachers College Press.

Jones D (1931) 'On phonemes' Travaux du Cercle Linguistique de Prague 4 : 74-79.

Jones D (1956) The Pronunciation of English Cambridge University Press.

Ladefoged P (1993) A Course in Phonetics (3rd edition) Fort Worth: Harcourt Brace.

Prator C & Robinett B (1985) A Manual of American English Pronunciation (4th edition) New York: Holt, Rinehart & Winston.

Roach P (1991) English Phonetics and Phonology (2nd edition) Cambridge University Press.

Wells J (1982) Accents of English (3 volumes) Cambridge University Press. [See J3 article.]

- (1990) Longman Pronunciation Dictionary Longman.

Back to the top.