International Journal on Studies in English Language and Literature (IJSELL)
Volume 5, Issue 10, October 2017, PP 36-46
ISSN 2347-3126 (Print) & ISSN 2347-3134 (Online)
http://dx.doi.org/10.20431/2347-3134.0510006
www.arcjournals.org
How Many Phonemes Does the English Language Have?
Aldo Luiz Bizzocchi
Universidade de São Paulo, Núcleo de Apoio à Pesquisa em Etimologia e História da Língua Portuguesa, R.
Dr. Albino Rodrigues de Alvarenga, 80, São Paulo SP – Brazil
*Corresponding Author: Aldo Luiz Bizzocchi, Universidade de São Paulo, Núcleo de Apoio à Pesquisa
em Etimologia e História da Língua Portuguesa, R. Dr. Albino Rodrigues de Alvarenga, 80, São Paulo SP
– Brazil
Abstract: Most phonology textbooks claim that the phonological system of the English language is composed
of 44 phonemes, of which 24 are consonants (actually, two are semivowels) and 20 are vowels. Yet, this
number results of a misinterpretation of the English vowel system, since several authors consider clusters of
sounds (diphthongs and pseudo triphthongs) as single phonemes, as well as combinatorial allophones of the
same phoneme as distinct phonemes. By employing the definitions of phoneme, phone, and allophone,
together with the criteria of commutation and substitution formulated by Trubetzkoy and the phonologists of
the Prague School, allied to the concept of neutralisation and archiphoneme of the traditional structural
phonology, it is possible to prove that the real number of phonemes of English is actually 35. This paper
intends to revise the English phonological system, especially regarding vowels, by using the same
methodology employed to describe the phonology of other languages, such as French, for example. As a
result, we will try to demonstrate that the current phonological description of the English language has a lot
of idiosyncratic and, therefore, a new view of this system will reflect on the phonemic transcription of words
made by dictionaries, making it simpler.
Keywords: English phonology; English vowel system; English vocalism; number of phonemes; allophone;
neutralisation; archiphoneme; Received Pronunciation; General American; English varieties.
Abbreviations: AmE: American English
BrE: British English
Fr.: French
GA: General American
It.: Italian
NE: New English
OE: Old English
Port.: Portuguese
RP: Received Pronunciation
1. INTRODUCTION
The number of English phonemes varies from one variety to another, but in general the language is
considered to have between 22 and 24 consonants (the difference is due to two units that in some
systems are considered phonemes and in others not), two semivowels and 20 vowels in the British
variety (BrE). In the American variety (AmE), 15 to 19 vowels are usually considered. Only this data
suffices to show that, if the number of consonants seems to be a point of agreement among specialists,
the determination of the number of vowels is subject to the individual interpretation of the researcher.
In fact, there are several possible criteria to individualise the distinctive units of a language, from
those more strictly linked to the phonetic aspect to those of a purely functional nature.
In any case, the spirit of phonological analysis, as formulated by its early researchers, such as Nikolai
Trubetzkoy and his colleagues in the Prague Linguistic Circle, excels at simplicity (of all possible
descriptions, one should always choose the simplest) and intuitiveness (the phoneme must be an
abstraction of the sounds of speech).
However, the usual description of the English phonological system, which includes 44 phonemes,
considers as single vowel phonemes several vowel clusters (diphthongs and, in certain cases, even
pseudotriphthongs), as well as treats as distinct phonemes what can be easily proved to be
combinatorial variants (allophones) of the same phoneme.
English has precisely the peculiarity of making a single vowel or semivowel sound in different ways
depending on the context where it lies. The tendency of the predominant phonological descriptions is
International Journal on Studies in English Language and Literature (IJSELL)
Page | 36
How Many Phonemes Does the English Language Have?
to treat each unit of phonetic realisation (i.e. each phone) with its respective context of occurrence as
an indivisible unit. For example, the cluster [eɪ̯] of late, although formed by two phones, [e] and [ɪ̯], is
taken as a single vowel phoneme, which would be opposed to [ɛ] of let and [ɛə̯] (or even [ɛə̯˞]) of lair.
The perspective of analysis proposed here increases very little the number of combinatorial
allophones (what is, by the way, perfectly reasonable, since the number of allophones, including free
and combinatorial ones, is naturally large due to the plethora of regional and national varieties of the
language), but reduces the number of phonemes, restricting them to the really distinctive minimal
units, thus to units represented at the level of the concrete realisation by a single phone. This
perspective represents a great economy in terms of the linguistic cost vs. yield relation, which is the
very spirit of the phonological description of a language, besides making it much simpler.
But before proceeding, we should recall that the analysis of the signifier of the linguistic sign – as
well as the analysis of every linguistic element – takes place at three levels, as defined by Coseriu [1],
namely, system, norm, and speech. At the level of concrete speech what we have is sound as an
acoustic phenomenon, only analysable and describable by means of graphs produced by
spectrograms. Therefore, when a linguist analyses the sounds of speech, what he or she really studies
are not the endless and unrepeatable concrete acoustic realisations that speakers produce in their dayto-day life, but abstractions made from these realisations, that is, patterns of pronunciation, units of
norm called phones, which, although numerous, are not infinite like concrete sounds, physically
detectable and recordable, and which, being units of norm, vary according to factors such as the
speaker‘s place of origin, age, sex, and schooling, among other variables.
In turn, at the level of the system, we can say that all these phones, some combinatorial (that is,
dependent on the phonological context where they lie), others free (that is, dependent only on the
factors mentioned above), are distributed in a number of minimal units distinguishing meanings that
are the phonemes, the number of which in principle must be the same for all speakers of the language
(or at least for all speakers of a particular national variety of the language).
The standard test to determine whether or not two phones represent the same phoneme is the so-called
commutation vs. substitution test: if the exchange of one phone by another, maintaining all other
phones of the word signifier, in all words in which both can occur, does not cause a change of
meaning, we say that these two phones are allophones of the same phoneme, in which case the
exchange is called substitution. Likewise, if two phones never occur in the same phonological
contexts, that is, they have complementary distribution, we also say that they are allophones of the
same phoneme.
When the exchange of a phone for another in the same conditions described above causes the change
of meaning, we say that these phones represent two distinct phonemes and, therefore, constitute a
minimal pair. In this case, the exchange is called commutation. For example, [f] and [v] are the normlevel realisations of the distinct phonemes /f/ and /v/ because fat and leaf mean things other than vat
and leave, respectively. In this sense, <fat> and <vat>, as well as <leaf> and <leave>, constitute
minimal pairs.
When two or more phones are usually opposed (i.e. represent distinct phonemes) in most contexts, but
in some of them this opposition does not take place, we say that there is neutralisation of the
phonemes that they represent. For example, we know that [s] and [ʃ] represent different phonemes
because see and ass are different from she and ash. However, the exchange of the one by the other
does not change the meaning when they are in final position of word and the next word begins with
[j], as in I miss you, which can be pronounced both as [aɪ̯ mɪs juː] and as [aɪ̯ mɪʃ juː] (especially in the
USA). The result of neutralisation is the emergence of an archiphoneme, a unit that brings together
the distinctive features of the neutralised phonemes. In the above example, the neutralisation of /s/ and
/ʃ/ results in archiphoneme /S/.
2. THE ENGLISH PHONOLOGICAL SYSTEM IN THE MAINSTREAM APPROACH
Table 1 shows the 44 English phonemes as defined in what I will henceforth call the ‗mainstream
approach‘, that is, the predominant one in current English phonology textbooks. The inventory of
phonemes is based on the two main varieties of English, the British (BrE), based on the Received
Pronunciation (RP), and the American (AmE), based on General American (GA). The representation
of the phonemes in this table uses the two usual systems: one, used in most American dictionaries; the
other, used in most British dictionaries.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 37
How Many Phonemes Does the English Language Have?
Table1. The English phonological system according to the mainstream approach
GA
Consonants and semivowels
RP Allophones
Context
word onset, before stressed
vowels
path,
campaign
[p]
after /s/, end of syllable
spend,
gap
[t , t̺ ] [2],
[3]
[t, t̺ ] [2], [3]
word onset, before stressed
vowels
after /s/, end of syllable
after a stressed vowel and
before an unstressed vowel
(especially GA)
word onset, before stressed
vowels
[ph] [2], [3]
/p/
/t/
GA
/ă/
/p/
h
/t/
Example
h
[ɾ]
RP
/a/1
stop, hat
diphthongs /aj/
and /aw/
by, out
/â/
/eə/
[ɛə̯, GA ɛ˞]
before /r/ and /R/
fairy, where
/ä/
/ɑː/
/ĕ/
/e/
[ɑː, GA ɑ, ɑ˞]
[ɛ]
[e]
all contexts
all contexts
in diphthong /ej/
father, star
red, play
play
/ē/
/iː/
[iː, GA i]
except before /r/
and /R/
fee, clean
better, petal
/b/
[b]
all contexts
bad, dab
[d, d̺]
all contexts
dab, bad
/d/
/d/
[ɾ]
after a stressed vowel and
before an unstressed vowel
(especially GA)
ladder, pedal
[ɡ]
all contexts
gap, log
/ĭ/
/ɪ/
/g/
[a]
play, state
/b/
/g/
by
except before /r/
and /R/
all contexts
after /s/, end of syllable
diphthong /aj/
[eɪ̯ , ɛɪ̯ ]
[tʃ, tʃw]
[k]
[ɑ]
/eɪ/
/tʃ/
/k/
[ɪ]
[i]
/j/
/f/
/th/
/dʒ/
/f/
/θ/
[dʒ, dʒ ]
/f/
[θ, tθ]
all contexts
all contexts
all contexts
job, ridge
face, staff
thing, both
/ī/
/î/
/ŏ/
/aɪ/
/ɪə/
/ɒ/
[aɪ̯ , ɑɪ̯ ]
[ɪə̯]
[ɒ, ɔ, GA ɑ]
/s/
/s/
[s]
all contexts
soap, stress
/ō/
/əʊ/
[oʊ̯, əʊ̯]
/sh/
/ʃ/
[ʃ, ʃw]
all contexts
shell, wash
/ô/
/ɔː/
[ɔː, RP oː, GA
ɔ]
/h/
/h/
[h]
[h, ɦ]
word onset, after consonants
between vowels
have
behave
/oi/
/ɔɪ/
[ɔɪ̯ ]
/v/
/v/
[v]
all contexts
vat, cave
/͝oo/
/ʊ/
[ʊ]
/dh/
/ð/
[ð, dð]
all contexts
then,
other, booth
/͞oo/
/uː/
[uː, ʉː, GA u]
w
/z/
/z/
[z]
all contexts
zebra, gaze
/ŭ/
/ʌ/
[ʌ, ɐ]
/zh/
/ʒ/
[ʒ, ʒw]
all contexts
genre, vision
/ū/
/aʊ/
[aʊ̯, æʊ̯]
/m/
/m/
[m]
all contexts
map, dam
/û/
/ɜː/
[ɜː, GA ɝ, ɚ]
/n/
/n/
[n]
all contexts
nap, man
—
/ʊə/
[ʊə̯]
/ng/
/ŋ/
[ŋ]
only after vowels
hang
/ə/
/ə/
[ə, ɨ]
[l]
before vowels (especially RP)
/l/
/l/
[ɫ] [4], [5]
all contexts
/r/
/r/
[ɹ, ɹw, r, ɾ] before vowels (RP), all contexts
[6]
(GA)
[ɹ, ɹ, ˞]
(/wh/) (/hw/) [hw, ʍ, w]
(/kh/) (/x/)
/y/
/w/
/j/
/w/
cat, out
/ā/
/ch/
/k/
Example
tap, contain
cat,
because
scan,
lack
c heap, watch
[kh] [2], [3]
Vowels
Allophones
Context
except in
[æ]
diphthong /aj/
all contexts
pit
before vowels and geology,
in final position
happy
except before /R/
by
before /r/ and /R/ fear, fearing
except before /R/
pot, sorry
except before /r/
and /R/2
low
all contexts
caught, all
except before /r/
and /R/
except in word
onset
except before /r/
and /R/
except in word
final
except before /R/
toy
foot, pull
food, super
hut, undo
out
furry, hurt,
all contexts
hors-doeuvre
before /r/ and /R/ pure, fury
unstressed
about, super,
syllable
apron
low
fall, felt
red, pray
star, start
after vowels (GA)
word onset
[x]
only in foreign words
[j]
[w]
before vowels
before vowels
what
loch,
chutzpah
yes
well
1
The vowel /a/ tends to be pronounced as [a] by many British speakers nowadays even in contexts other than
the diphthongs /aj/ and /aw/. Conversely, many American speakers tend to pronounce [æ] closer and tenser,
making sometimes the distinction between [æ] and [ɛ] (that is, between /a/ and /e/) difficult to perceive.
2
In AmE, /ō/ can also occur before /r/, as in four [foʊ̯ɹ].
International Journal on Studies in English Language and Literature (IJSELL)
Page | 38
How Many Phonemes Does the English Language Have?
The alleged phonemes /hw/ and /x/ have been included in Table 1 because they are sometimes
considered to be an integral part of the English phonological system. However, /hw/, even when
pronounced as a single phone [ʍ] (a voiceless labiovelar fricative [7], [8], [9] or approximant [10]), is
phonologically analysed as the cluster of phonemes /h/ and /w/. In favour of this position is the fact
that most of the speakers who distinguish between the pronunciation of whine and wine realise <wh>
as [hw] and not as [ʍ]. In addition, historically <wh> comes from <hw> (OE hwæt > NE what). A
similar phenomenon occurs with the pronunciation of <hu> in huge [çʉːdʒ], which is nonetheless
phonologically analysed as /hjuːʤ/ [11].
In turn, the voiceless velar fricative [x] occurs in words such as loch, which is usually pronounced
[lɒx] in Hiberno-, Scottish, South African and Welsh English – but as [lɒk] in the other varieties.
Nevertheless, loch is a loanword from Scots, which, strictly speaking, is another language, although
many consider it a dialect of English. In actuality, a dialect is a non-standardised language, but still a
language, therefore something different from a simple variety. In this sense, British English and
American English are varieties of English, just as Californian and New Yorker are varieties of
American English. In contrast, Scots is a language independent of English, with its own vocabulary,
grammar, and spelling, and today, already having its standardised written form, it cannot even be
considered as a mere dialect any longer.
Now if loch is strictly a foreign word, pronouncing it as [lɒx] is the same as pronouncing cordon bleu
as [kɒʁˈd ɔ̃ blø]: these are words that, although they are part of the English lexicon, are not fully
nationalised and, therefore, many speakers pronounce them in a foreign fashion.
To this list of phonemes, we can add the archiphonemes presented in Table 2, in which the symbol
represents neutralisation and Ø stands for the zero phoneme, that is, the absence of sound.
Table2. English vocalic and consonantal archiphonemes
Archiphoneme
/E/ = /ə/ Ø [12], [13], [14]
/I/ = /ə/ /ɪ/
/E‘/ = /E/ /I/ = /ə/ /ɪ/ Ø
/R/ = /r/ Ø
/T/ = /t/ /ʧ/ Ø
/D/ = /d/ /ʤ/ Ø
/S/ = /s/ /ʃ/
/Z/ = /z/ /ʒ/
Context of neutralisation
In unstressed syllable: battery, wooden
In unstressed syllable: animal
In unstressed syllable: ordinary, Latin
In word-medial position:
/r/ or Ø before consonants: part
In word-final position:
/r/ or Ø before consonants: far from
/r/ before vowels: far away
In word-final position:
/t/ or /ʧ/ before /j/: got you
/t/ or Ø after /n/ and before vowels: in front of
/t/ before the other phonemes and in absolute final position
In word-final position:
/d/ or /ʤ/ before /j/: and you
/t/ or Ø after /n/ and before vowels: on and on
/d/ before the other phonemes and in absolute final position
In word-final position:
/s/ or /ʃ/ before /j/: miss you
/s/ before the other phonemes and in absolute final position
In final position:
/z/ or /ʒ/ before /j/: is you
/z/ before the other phonemes and in absolute final position
Finally, the mainstream approach usually considers the existence of the triphthongs /eɪə/, /aɪə/, /ɔɪə/,
/aʊə/ and /əʊə/ (or /oʊə/). Actually, these clusters are best analysed as sequences of two phonemes
(one diphthong plus /ə/). In our reinterpretation of this system, we will see that, in fact, these are three
phonemes, although in certain varieties some of them may sound like hiatuses or even monophthongs:
layer [ˈleɪ̯ə(ɹ)], [ˈlɛɪ̯ə(ɹ)], [ˈlɛ.ə(ɹ)], [ˈlɛə̯(ɹ)], [ˈlɛːɹ].
3. REINTERPRETING THE MAINSTREAM SYSTEM
As we said, there is no objection to English consonantism; my disagreement refers to vocalism and
semivocalism. Therefore, the phonological description proposed here maintains the consonant
phonemes of the mainstream system, but reanalyses vowels and semivowels.
Vocalism in the BrE variety
British English actually has 11 vowel phonemes, six short and five long, namely (Table 3):
International Journal on Studies in English Language and Literature (IJSELL)
Page | 39
How Many Phonemes Does the English Language Have?
Table3. British English vowel phonemes in the new approach
Short vowels
Allophones
[ɛ]
[e]
red, play
play, state
[ɪ]
all contexts
pit
[ɒ, ɔ]
[ɔ]
before vowels and in
final position
except before /R/
in diphthong /oj/
geology,
happy
pot, sorry
voice
[ʊ]
all contexts
foot, pull
[u, ʉ]
before vowels
[ʌ, ɐ]
all contexts
influence
hut, hurry,
undo
animal, apron
go, low
Examples
yes
[ɑ]
[a]
/e/
/i/
[i]
/o/
/u/
/ʌ/
Semivowels
[ə, ɨ]
[ə]
Allophones
[j]
/j/
3
[ɪ̯ ]
[ə̯]
/w/
Examples
except in diphthong
/aj/
diphthong /aj/
diphthongs /aj/ and
/aw/
all contexts
in diphthong /ej/
[æ]
/a/
Contexts
[w]
[ʊ̯]
unstressed syllable
diphthong /ʌw/
Contexts
before vowels
after vowels, except
before /r/ and /R/
after vowels, before
/r/ and /R/
before vowels
after vowels
Long
vowels
Allophones
Contexts
Examples
/aː/
[ɑː]
all contexts
father, star
[iː]
except before /r/
and /R/
fee
[ɪə̯]
before /r/ and /R/
fear, fearing
[ɔː, oː]
all contexts
caught, all, sport
[uː, ʉː]
except before /r/
and /R/
before /r/ and /R/
cat, out
by
by, out
/iː/
/oː/
/uː/
/ʌː/
[ʊə̯]
[ɜː]
all contexts (most
frequent before /r/
and /R/
food, super
pure, fury
furry, hurt, horsdoeuvre
play, pain
fairy, their
wine, queen
out, low
Vowel Clusters
As we can see, the phonetic diphthongs [aɪ̯], [eɪ̯], [ɔɪ̯], [aʊ̯], [əʊ̯], [ɛə̯], [ɪə̯], and [ʊə̯] are not phonemes
but clusters of a vowel phoneme and a semivowel. Fundamentally, the six English short vowels
present different allophones as they occur alone or followed by /j/ or /w/. These glides, in turn, also
present different allophones as they are before ([j] or [w]) or after the vowel ([ɪ̯] or [ʊ̯]), forming,
therefore, a rising or falling diphthong. The semivowel /j/ also has an allophone [ə̯] in coda position
when followed by the phoneme /r/ (pronounced [ɹ]) or the archiphoneme /R/, which neutralises /r/ and
Ø.
How to prove it? Let be the words let [lɛt], lay [leɪ̯], and lair [lɛə̯(ɹ)]. Hence we assume, in the
mainstream approach, the alleged phonemes /ĕ/ (or /e/), /ā/ (or /eɪ/), and /â/ (or /eə/). It is easy to see
that the phonetic sequence [ɛə̯] only occurs before /r/ and /R/ (*/leət/and */leə/ are impossible in
English), while /e/ and /eɪ/ never occur before /r/ and /R/ (*/le (ɹ)/4 and */leɪ(ɹ)/ are equally
impossible). We say that /eɪ/ and /eə/ have complementary distribution, therefore, they are, in
principle, combinatorial allophones of the same phoneme. Incidentally, there is a historical
explanation for this. At some point in the evolution of Middle English, the free a became long and
then began to move forward and progressively diphthongise, merging with the diphthong ai. We then
had something like /a/ > /aː/ > /æː/ > /ɛː/ > /ɛɪ/ > /eɪ/. One can imagine that at some point Mary and
fairy sounded [ˈmɛɪ̯ɹi] and [ˈfɛɪ̯ɹi]. As the retroflex [ɹ] of English causes the lowering of the previous
vowel as a matter of phonetic harmony or assimilation, the semivowel [ɪ̯] lowered to [ə̯], a fact for
which undoubtedly contributed the neutralisation of /ɪ/ and /ə/ in unstressed position. This means that
every /eə/ was once /eɪ/.
If, therefore, these two diphthongs have complementary distribution, /e/ and /eɪ/, on the contrary, are
opposed, for let [lɛt] is different from late [leɪ̯t]. However, if we change [ɛ] for [e] and vice versa, we
seem to have no change of meaning (the pronunciations [let] of let and [lɛɪ̯t] of late do exist,
3
The vowel [ɨ] can equally be considered as an unstressed allophone of /i/; however, I prefer to consider it an
allophone of /ʌ/, since, in the contexts where [ɨ] occurs, [ə] occurs as well, but not always [ɪ]. In contrast, in the
contexts where [ɪ] alternates with [ɨ], it also alternates with [ə], so that [ɨ] seems to be distributionally closer to
[ə] than to [ɪ].
4
Strictly speaking, there is in the USA the pronunciation [lɛɹ] as a variant of [lɛə̯ɹ], that is, it is possible to
establish an archiphoneme that neutralises [ɛ] and [ɛə̯] before /r/ (but not before /R/) in AmE.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 40
How Many Phonemes Does the English Language Have?
especially in Great Britain and Australia, although in certain places only [ɛ] is used and in others only
[e]). Yet, the most common in both RP and GA is that [e] is only found in the cluster [eɪ̯] and that [ɛ]
does not occur followed by [ɪ̯]. It follows that /eɪ/ is not a phoneme, but the combination of /e/ and the
semivocalic /ɪ/, which also occurs in /aɪ/ and /ɔɪ/. Next, we find that [j] and [ɪ̯] have complementary
distribution, since [j] only occurs before vowel and [ɪ̯], only after vowel. That is, [j] and [ɪ̯] are
combinatorial allophones of the same phoneme /j/ as they are in onset or coda position in the syllable.
It follows that [eɪ̯] does not correspond to one phoneme, but two: /e/ and /j/. It also follows that the
cluster [ɛə̯], which we had proved to be a variant of [eɪ̯], is not a phoneme, but the same combination
/e/ + /j/, where the phoneme /j/ sounds like [ə̯] before /r/ and /R/5. In short, we have the following
phonemes and their allophones:
/e/ [ɛ] and [e] (free allophones in some English varieties and combinatorial in others);
/j/ [j], [ɪ̯], and [ə̯] (combinatorial allophones).
The same reasoning applied to /j/ and its allophones is true for /w/ and its allophones [w] (in onset
position) and [ʊ̯] (in coda position).
As for the other diphthongs, the same glides /j/ and /w/ can be abstracted, alongside with the short
vowels /a/, /o/, and /ʌ/, with their allophones [a], [æ], [ɑ], [ɒ], [ɔ], [o], [ʌ], [ɐ], [ə], and [ɨ].
On the diphthong [əʊ̯] of BrE, we conventionally represented it phonologically as /ʌw/, but the fact
that it is pronounced as [oʊ̯] in other varieties of English, including some of Great Britain, could allow
us to represent it as /ow/. In this case, [ə] would be the allophone of /o/ before the semivowel /w/.
However, since [ə] is an allophone of /ʌ/, we chose the simplest representation.
As for the alleged triphthongs of the mainstream approach, we actually have the combination of a
diphthong plus /ə/, hence three phonemes: /ajʌ/, /ejʌ/, /ojʌ/, /awʌ/, /ʌwʌ/ (/owʌ/ in AmE). On the
representation of [ə] as /ʌ/, we will speak later on (see item Schwa).
Vowels i and u
In relation to /i/ (i.e. short i), we have two allophones: [ɪ], which is the ‗normal‘ realisation of the
vowel, and [i], which occurs in unstressed position before another vowel (e.g. in geology [dʒiˈɒlədʒi]
and in absolute final position, as in geology, happy, etc.
In final position, the mainstream approach sometimes considers [i] a phoneme distinct from [ɪ], using
to support such a position minimal pairs such as studded [ˈstʌdɪd] vs. studied [ˈstʌdid] and taxes
[ˈtæksɪz] vs. taxis [ˈtæksiz]. However, the postulation of a phoneme that occurs only in final position,
followed or not by /d/ or /z/ is quite uneconomical (not to say unreasonable) in terms of phonological
yield (cost of opposition vs. number of minimal pairs). Moreover, the pronunciations [dʒɪˈɒlədʒɪ],
[ˈhæpɪ], [ˈstʌdɪ], and [ˈtæksɪ] also occur [15], and the substitution of [i] for [ɪ] in these cases does not
entail any change of meaning. In pop music lyrics, by the way, it is common to hear a word like baby
pronounced [ˈbeɪ̯be], in which the final [e] is an exaggeratedly open manner of pronouncing [ɪ].
Kreidler [16], McCully [17] and Roach [18] suggest that [i] occurs in contexts where the
neutralisation of /ɪ/ and /iː/ takes place, and, thus, is an archiphoneme. However, in final position,
especially in the BrE variety, it cannot be said that there is such a neutralisation because there are a
number of words (caries, series, species, etc.) in which the ending /iːz/ sounds different from /iz/
(compare caries vs. carries or chickaree vs. chicory) [19], [20].
In addition, it should be noted that the ending [iz] occurs only in the plural of nouns ending in i, ie, or
y (taxis, calories, theories) and in the third person singular present indicative of verbs ending in y
(carries, copies, etc.). The ending [id] only occurs in the past and participle of regular verbs ending in
y (carried, copied, etc.). The pronunciation of the plural or third person singular ending -es after
sibilant consonant (/s/, /z/, /ʃ/, /ʧ/, /ʤ/) is always [ɪz], as well as the pronunciation of the
past/participle ending -ed after dental stop (/t/, /d/) is always [ɪd]. That is, the pronunciations [iz] and
As a result, the neutralisation [ɛ] [ɛə̯] referred to in the previous footnote can be formally translated as /e/
/ej/ / __/r/, that is, there would be in AmE an archiphoneme /J/ = /j/ Ø so that the word lair would be
transcribed as /leJr/. In BrE, the transcription would be /lejR/, that is to say, in Britain it is optional to realise /r/,
whereas in the USA it is the realisation of yod (in its allophone [ə̯]) which is optional.
5
International Journal on Studies in English Language and Literature (IJSELL)
Page | 41
How Many Phonemes Does the English Language Have?
[id] seem to occur by analogy to the forms ending in [i], of which they are inflections. In fact, the
pronunciations [ˈkɒpɪ], [ˈkɒpɪz] are also found and there is no change of meaning in them.
We can say, therefore, that the oppositions studded vs. studied and taxes vs. taxis are not of
phonological but morphological nature, since they correspond to a different segmentation of the
morphemes: studd-ed /ˈstʌd-id/ vs. studi-ed /ˈstʌdi-d/, tax-es /ˈtaks-iz/ vs. taxi-s /ˈtaksi-z/. In this case,
we would say that the allophone [i] occurs in unstressed end of morpheme and [ɪ] in the other
contexts.
In short, the phoneme /i/ presents the allophones [ɪ] and [i], which are not to be confused with the
phoneme /iː/. Analogously, the phoneme /u/ has the allophones [ʊ] and [u] (which can also be
pronounced [ʉ]) in the same contexts: [u] or [ʉ] before vowel (influence, usual) and [ʊ] in other
contexts, except in final position, where only /uː/ occurs (residue).
The long vowels /iː/ and /uː/ have their respective allophones [ɪə̯] and [ʊə̯] before /r/ and /R/. The
reason is the previously mentioned vowel lowering caused by /r/. The diphthongisation of /iː/ and /uː/
is a purely phonetic process, but we can also analyse it from the phonological point of view, since, in
a certain way, /iː/ = /ij/ and /uː/ = /uw/6. In this sense, there would have been a lowering of /j/ or, in
phonetic terms, the lowering of [ɪ̯] to [ə̯]: /iː/ = [ij] > [iɪ̯] > [iə̯] > [ɪə̯].
Schwa
English has the vowel /ʌ/, which comes from an old short u that has been kept in a few words (full,
push) and in the others has evolved into a near-open back sound. Currently, many speakers realise /ʌ/
as a central near-open vowel [ɐ], although the original pronunciation [ʌ] remains alive [21]. Although
this vowel, like the other full vowels of English, occurs predominantly in stressed syllables, it can also
occur in unstressed syllable (undo). In turn, schwa (/ə/) occurs, in the BrE variety, exclusively in
unstressed syllables and is the result of the reduction of full vowels in unstressed syllable. In fact, in
unstressed position, all vowels converge to [ə], [ɪ] or [ɨ] (less often to [ʊ]). In this sense, cases like
undo would be examples of secondary stress (therefore, the syllable un of undo would not be totally
unstressed). Ladefoged [22] and Bolinger [23], however, consider this as a mere difference in vowel
quality rather than stress. That is the position also taken here.
Anyway, in non-stressed syllables, [ʌ] and [ə] (and its variant [ɨ]) are not opposed, since [ənˈdʉː] and
[ʌˈbaʊ̯t] do not mean things other than [ʌnˈdʉː] and [əˈbaʊ̯t]. Given that [ə] and its variants never
occur in stressed syllable in the BrE variety, it can be said that [ʌ] and [ə] have complementary
distribution and, therefore, they are allophones of the same phoneme. Thus, we posit the phoneme /ʌ/,
whose realisation in stressed syllable is [ʌ] or [ɐ] (free variants) and in unstressed syllable can be [ə]
or [ɨ] (in cases where it may sound [ɨ] – usually in the posttonic syllable – it is neutralised with the
phoneme /i/, producing the already mentioned archiphoneme /I/).
On the other hand, the phone [ɜː], which occurs in general before /r/ and /R/ (but also in other
contexts, as in hors-doeuvre and cordon bleu), is potentially opposed to [ʌ] in stressed syllable (hurry
[ˈhʌɹi] vs. furry [ˈfɜːɹi]) and actually in unstressed syllable (forward [ˈfɔːwəd] vs. foreword
[ˈfɔːwɜːd]). As we already know that [ʌ] and [ə] are allophones of /ʌ/, the opposition, both in stressed
and unstressed syllable, occurs between [ɜː] and /ʌ/. For simplicity and to maintain the symmetry
between short and long vowels, we say that [ɜː] is the long counterpart of /ʌ/ and represent it
phonologically as /ʌː/.
In AmE, as we shall see later on, there is a tendency to pronounce [ə], [ʌ], and [ɜː] always as [ə].
Vocalism in the AmE Variety
Now let us look at the English vocalism in the American variety. In the USA, the main opposition is
not between long and short vowels, but between tense and lax vowels. Strictly speaking, since vowel
quantity (i.e. duration) is not a relevant distinctive feature, the distinction between vowels is based on
their timbre. Therefore, the denomination tense vs. lax does not really have to do with tension or
intensity, but with degree of opening and place of articulation. Therefore, I will replace these terms
with close and open, respectively. Let us look at Table 4.
6
Proof of this is that /iː/ and /uː/ are pronounced as [ɪj] and [ʊw] in many places. In some of them, by the way,
these diphthongs are already evolving to [ej] and [ow], respectively.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 42
How Many Phonemes Does the English Language Have?
Table4. American English vowel phonemes in the new approach
Open vowels
Allophones
Contexts
Examples
/æ/
[æ]
cat, out
/ɛ/
[ɛ]
except in
diphthong
/aj/
except in
diphthong
/ej/
in diphthong
/ej/
except before
vowels
before /r/
except in
diphthong
/ow/
in diphthong
/ow/
except before
vowels
before /r/
[e]
/ɪ/
[ɪ]
/ɔ/
[ɪə̯]
[ɔ]
[o]
/ʊ/
[ʊ]
[ʊə̯]
Close
vowels
/ɑ/
Allophones
Contexts
Examples
[ɑ]
all contexts
father, star,
by, pot,
sorry
/i/
[i]
except
before /r/
and /R/
fee,
geology,
happy
/u/
[uː, ʉː]
food, super
/ə/
[ə, ʌ, ɜ]
except
before /r/
and /R/
stressed
syllable
red, there
play, state
pit, pin
fear,
fearing
caught,
all, sport,
soft, voice
go, low,
four7
foot, pull
pure, fury
[ə, ɨ]
Semivowels
/j/
Allophones
[j]
[ɪ̯ ]
/w/
[w]
[ʊ̯]
Contexts
before
vowels
after vowels,
except before
/r/ and /R/
before
vowels
after vowels
unstressed
syllable
hut, hurry,
furry, hurt,
horsdoeuvre
undo,
animal,
apron
Examples
yes
play, pain
wine,
queen
out, low
In AmE, the vowels [ə], [ʌ], and [ɜ] have complementary distribution: [ə] in unstressed syllables, [ʌ]
in stressed syllables except before [ɹ], and [ɜ] in stressed syllables before [ɹ] [24]. They are, therefore,
allophones of a single phoneme /ə/. By the way, many American speakers pronounce [ə] in all three
cases.
In BrE, only the long counterpart of /e/ (i.e. */eː/) is missing; the other short vowels have their long
counterparts. In AmE, in addition to long /e/, the close counterpart of /ɔ/ is missing (since [o] is the
allophone of /ɔ/ in the diphthong [oʊ̯]) and also the open counterpart of /ə/. What happens is that,
unlike BrE, AmE does not distinguish between [ɑː] and [ɒ] (it is the so-called father–bother merger),
pronouncing both as [ɑ]. In contrast, many words that in BrE sound with [ɒ] (for example, soft and
song) sound in AmE with [ɔ]. In other words, the British phonemes /aː/, /o/, and /oː/ undergo a
rearrangement in AmE, regrouping into /ɑ/ and /ɔ/: the British phoneme /o/ is treated in AmE
sometimes as /aː/ (i.e. /ɑ/) and sometimes as /oː/ (i.e. /ɔ/).
Similarly, the British vowel /aː/ is often converted into the American /æ/: banana BrE [bəˈnɑːnə]
(phonological transcription /bʌˈnaːnʌ/) vs. AmE [bəˈnænə] (phonological transcription /bəˈnænə/ or
/bʌˈnănʌ/).
On the latter transcription, incidentally, if we want to unify the phonological representation of BrE
and AmE varieties, using the same phonetic symbols for both (which would be desirable when it
comes to the same language), we can establish the following convention:
7
The word four (and others like it) admits both pronunciations [foʊ̯ɹ] (sometimes, just [foɹ]) and [fɔɹ]. See
footnote 2.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 43
How Many Phonemes Does the English Language Have?
Table5. Equivalence of phonemes between British and American varieties of English
Phoneme
/ă/
/ā/
/ĕ/
/ĭ/
/ī/
/ŏ/
/ō/
/ŭ/
/ū/
/̆ʌ/
/͞ʌ/
BrE (RP)
/a/
/aː/
/e/
/i/
/iː/
/o/
/oː/
/u/
/uː/
/ʌ/
/ʌː/
Allophones
[a], [æ], [ɑ]
[ɑː]
[ɛ], [e]
[ɪ], [i]
[iː], [ɪə̯]
[ɒ], [ɔ]
[ɔː]
[ʊ]
[uː], [ʉː], [ʊə̯]
[ʌ], [ə], [ɨ]
[ɜː]
AmE (GA)
/æ/
/ɑ/
/ɛ/
/ɪ/
/i/
Allophones
[æ], [a]
[ɑ]
[ɛ], [e], [ɛə̯]
[ɪ], [i], [ɪə̯]
[iː]
/ɔ/
/ʊ/
/u/
/ə/
[ɔ]
[ʊ], [ʊə̯]
[uː], [ʉː]
[ʌ], [ɜ], [ə], [ɨ]
As we can see, in AmE the diphthongs [ɪə̯] and [ʊə̯], in the varieties in which they occur before /r/, are
not allophones of /ī/ and /ū/ as in BrE, but of /ĭ/ and /ŭ/, respectively.
Unstressed Syllables and Vowel Reduction
English has a peculiar feature, which is vowel reduction in unstressed syllable. So much so that the
most frequent vowels in this position are [ə] and [ɪ], being [ɨ] a variant of both in contexts in which
they are neutralised. The phenomenon of reduction is visible when we compare a pair like photograph
vs. photography: in photograph, the first <o> and the <a> are full ([əʊ̯], [ɑː]) while the second <o> is
reduced [ə]; in photography, the first <o> and the <a> are reduced ([ə]) while the second <o> is full
([ɒ]). The phonological approaches that consider the existence of three degrees of stress in English
(primary stress, secondary stress and unstressed syllable) consider that the full vowels occur in
primary and secondary stresses and that, therefore, the vowel reduction is exclusive of unstressed
syllables. Ladefoged [25] considers that English has only stressed and unstressed syllables, the
secondary stress being a phonetic, not phonemic feature.
Considering the existence of vowel reduction could lead us to postulate a series of archiphonemes (for
example, /A/ = /a/ /ʌ/, /O/ = /o/ /ʌ/, etc.) to give account of each of these reductions. Nevertheless,
it is simpler to admit, as in many other languages (French, Italian, German, Portuguese) that certain
morphemes (e.g. phot- and graph-) admit allomorphs as they are stressed or unstressed (/fʌwt-/ vs.
/fʌt-/, /graːf-/ vs. /grʌf-/). Examples of this type in other languages are: Fr. mener [məˈne] vs. mène
[mɛn], It. prendo [ˈprɛndo] vs. prendiamo [prenˈdjamo], Port. chover [ʃuˈver] vs. chove [ˈʃɔvɨ].
The same applies to most of the monosyllables, which in English have two pronunciations, one
stressed and one unstressed: a ([eɪ̯] vs. [ə]), an ([æn] vs. [ən]), of ([ɒv] vs. [əv]), for ([fɔː] vs. [fə]), me
([miː] vs. [mɪ]), them ([ðɛm] vs. [ðəm]), etc.
4. CONCLUSION
For a number of reasons (tradition, idiosyncrasy, attachment to orthography), the phonological
description of English, especially with regard to vowels, has not followed the usual criteria of
phonological theory, as defined by its creators, Jan Baudouin de Courtenay, Mikołaj Kruszewski, Lev
Shcherba (1876-1877), Antoni Dufriche- Desgenettes (1873) and especially the linguists of the Prague
Linguistic Circle, notably Nikolai Trubetzkoy (Grundzüge der Phonologie, 1939), starting from the
distinction between langue (language) and parole (speech), proposed by Ferdinand de Saussure
(1916), which theory, by the way, has been little altered in its general principles by more recent
approaches (generative, auto segmental, evolutionary phonology, etc.).
The description of English vocalism currently in use considers vowel clusters composed of two
phonetic sounds (phones), or sometimes up to three, as a single vowel, as well as considers sounds or
clusters of sounds having complementary distribution as if they were distinct phonemes. In this paper,
I tried to show that the elimination of these methodological inconsistencies leads to a much simpler
description of English, with a smaller number of vowel phonemes, and in which each vowel or
semivowel is a phoneme, that is, phonetic diphthongs correspond to phonological diphthongs and not
to single phonemes.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 44
How Many Phonemes Does the English Language Have?
For example, in the current approach, [aɪ̯], [eɪ̯] and [ɔɪ̯] are seen as three phonemes, despite the
recurrence of the final glide [ɪ̯] and despite [a], [e] and [ɔ] occurring autonomously. In addition, the
phonological representation of these alleged phonemes as /aɪ/, /eɪ/ and /ɔɪ/ is inappropriate in that each
phoneme should be represented by a single International Phonetic Alphabet symbol. (For that reason,
incidentally, the phonological representation of the affricates [tʃ] and [dʒ] is done internationally with
the symbols /ʧ/ and /ʤ/ or, failing them, with /͡tʃ/ and /͡dʒ/.)
It is a fact that the pseudophonemes /aɪ/ and /eɪ/ come historically from the simple Old English vowels
ī and ā (/ɔɪ/ was borrowed directly from French in this form), but the fact that a current diphthong
comes from an old monophthong does not make necessarily this diphthong a single phoneme. Proof
of this is that the Latin long ē has evolved in modern French to <oi> (pronounced [wa]), but the
descriptions of the French phonological system do not consider /wa/ a phonemically indivisible unit.
In effect, /wa/ is phonologically opposed to /wɛ/, /wi/, etc., and also to /ja/, /la/, /da/, and so on, from
which two distinctive units are abstracted, /w/ and /a/.
The fact that the elements [a], [e] and [ɔ] of the pseudophonemes /aɪ/, /eɪ/ and /ɔɪ/ sound a bit different
from the autonomous vowels [æ], [ɛ] and [ɒ] does not mean much in phonological terms: the fact that
these autonomous monophthongs can also be pronounced as [a], [e] and [ɔ] in certain varieties of the
language, as well as the finding that in other varieties the series [a]/[e]/[ɔ] and [æ]/[ɛ]/[ɒ] have
complementary distribution (the first series only in diphthongs and the second only in
monophthongs), already prove that they are only combinatorial allophones of the same phonemes.
The finding that /eɪ/ and /eə/ (best represented as /eɪ̯/ and /eə̯/) also present complementary
distribution equally leads to the conclusion that they are not two distinct phonemes (actually, they are
not even phonemes). The same reasoning applies to all other vowels.
In short, disregarding these facts leads to a more complex phonological description and farther from
reality than it should be. I hope, therefore, to have been able to demonstrate in this article that the
phonological reality of the English language is much simpler than the available phonological
textbooks suggest.
The revision of English phonology, especially in relation to vocalism, will certainly imply the
reformulation of the descriptions found in the language dictionaries, which will facilitate not only the
understanding of the system by foreign speakers who are learning English as well as the native
speakers themselves, who sometimes need to consult the dictionary to learn the correct pronunciation
of a particular word, since the official spelling does not help much in these cases.
In this regard, although this study has not been exhaustive in terms of regional allophones and other
specifics of the phonetics and dialectology of the language, I hope to have made an important
contribution to the studies of phonetics and phonology of English.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
Coseriu, E. Teoría del lenguaje y lingüística general. 3rd ed. Madrid: Gredos, 1973, pp. 94-9.
Celce-Murcia, M.; Brinton, D.; Goodwin, J. Teaching pronunciation: a reference for teachers of English to
speakers of other languages. Cambridge: Cambridge University Press, 1996, pp. 62-7.
Roach, P. English phonetics and phonology: a practical course. 4th ed. Cambridge: Cambridge University
Press, 2009, pp. 26-8.
Wells, J. C. Accents of English. Cambridge: Cambridge University Press, 1982, p. 490.
Wells, op. cit., 1982, p. 550.
Ladefoged, P. A course in phonetics. 4th ed. Fort Worth: Harcourt College Publishers, 2001, p. 55.
Cruttenden, A. Gimson’s pronunciation of English. London: Hodder Arnold, 2008, p. 230.
McMahon, A. An introduction to English phonology. Edinburgh: Edinburgh University Press, 2002, p. 31.
Giegerich, H. English phonology: an introduction. Cambridge: Cambridge University Press, 1992, p. 36.
Ladefoged, P. Vowels and consonants. Chichester: Blackwell, 2006, p. 68.
Roach, op. cit., 2009, p. 43.
Roach, op. cit., 2009, pp. 100-1.
Kreidler, C. The pronunciation of English. Chichester: Blackwell, 2004, p. 84.
Wells, op. cit., 1982, p. 55.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 45
How Many Phonemes Does the English Language Have?
[15] Lewis, J. W. HappYl and Reconnoitred. The Home Page of Jack Windsor Lewis (http://www.yek.m
e.uk/happyland.html). Access on 21st Aug 2017.
[16] Kreidler, op. cit., 2004, pp. 82-3.
[17] McCully, C. The sound structure of English. Cambridge: Cambridge University Press, 2009, pp. 123-4.
[18] Roach, op. cit., 2009, pp. 66-8.
[19] Wells, J. C. Sounds interesting. Cambridge: Cambridge University Press, 2014, p. 53.
[20] Bolinger, D. Intonation and its parts: melody in spoken English. Redwood, CA: Stanford University Press,
1986, p. ―3‖ 48.
[21] Roca, I; Johnson, W. A course in phonology. Chichester: Blackwell, 1999, p. 135.
[22] Ladefoged, op. cit., 2001.
[23] Bolinger, op. cit., p. ―3‖ 51.
[24] Wells, op. cit., 1982, pp. 121, 132, 480.
[25] Ladefoged, op. cit., 2001, § 5.4; Preliminaries to linguistic phonetics. Chicago: University of Chicago
Press, 1980 p. 83.
AUTHOR’S BIOGRAPHY
Aldo Luiz Bizzocchi, holds a PhD in Linguistics from the University of São
Paulo with postdoctoral degrees in Comparative Linguistics and Etymology.
He is currently a researcher at the Centre for Research in Etymology and
History of the Portuguese Language at the University of São Paulo
(www.nehilp.org). He is the author of two books and co-author of five other
works. He is also devoted to the scientific popularisation of linguistics by
means of blogs, articles in magazines and videos on the internet.
Citation: Aldo Luiz Bizzocchi "How Many Phonemes Does the English Language Have?" International
Journal on Studies in English Language and Literature (IJSELL), vol 5, no. 10, 2017, pp.36-46.
doi:http://dx.doi.org/10.20431/2347-3134.0510006.
Copyright: © 2017 Authors. This is an open-access article distributed under the terms of the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium,
provided the original author and source are credited.
International Journal on Studies in English Language and Literature (IJSELL)
Page | 46