Graphemes, Phonemes & Allophones

Letters and Sounds (‘Text to Speech’)

In order to establish a systematic reference system in the first place, it is quite useful to start by having a look at letter-to-sound correspondences, only that in this context, we’ll refer to the former as graphemes and represent them in angled brackets (<>) and to the latter as phonemes, which we’ll enclose in forward slashes (//) or sometimes in square ([]) brackets, if we want to talk about particular realisations (or phones). This distinction – which will be clarified further below – is particularly important because we unfortunately tend to use the same terms, i.e. vowels & consonants, for both letters and sounds, so that occasionally, there may be grounds for misunderstanding. For example, when asked about the vowels of English, most native speakers would automatically respond by saying /eɪ/, /iː/, /aɪ/, /əʊ/, /juː/, of course meaning the names for <a>, <e>, <i>, <o>, <u>, because of the predominance of written language in our literate society. But of course, in terms of vowel sounds, this rendition does not actually even cover the simple vowels, which the single grapheme correspondences would seem to suggest. In order to shed some more light on what kinds of relationships exist between graphemes and phonemes, we will therefore look at the different types of letter/sound classes in some detail. Before we begin, though, it is important to bear in mind right from the start that we may not only be dealing with one-to-one correspondences, but that there may well also be one-to many-relationships, where a single grapheme/phoneme may have multiple realisations and vice versa.

  1. As an initial exercise, think about which sounds the vowel graphemes listed above may represent and maybe in which combinations, as usual trying to come up with a suitably transcribed list of words. Can you already draw some initial conclusions about ‘text-to-speech rules’ for vowels from this?
  2. Try to do the same thing for the consonant graphemes.

Once you’ve finished the exercises, we’ll start solving the above questions by first looking at the consonants because their realisation generally involves less variation, and will then proceed to the vowels.

Consonant Graphemes/Grapheme Combinations and Their Realisations

Below is a table of possible consonant graphemes or grapheme combinations. Where a potential doubling of graphemes may be possible without causing a gemination (i.e. a double articulation) of the phoneme, the second grapheme is given in round brackets; at other times the round brackets indicate additional optional elements.

<b(b)> /b/ bill, trouble, shabby, tabby, tab +
<bt> /t/
debt, doubt
<c> /k/
cat, chocolate, mac
special, ocean
<cc> /k/
accurate, occur
<(t)ch> /tʃ/
check, match
machine, mustache
chaos, Bach
<ck> /k/ chicken, back, thick +
<d(d)> /d/
/dʒ/, /dj/
dumb, middle, bad
gradual, residual
<f(f)> /f/
find, coffee, dwarf, shelf
<g(g)> /g/
give, guest, luggage, bag
George, engorge, huge, urge
rouge, beige
<gh> /g/
cough, laugh
eight, bough
<gn> /n/
sign, foreign
signal, signature
<h> /h/
home, behind
hour, Shah
<j> /dʒ/
John, joke, majority
<k> /k/ kilo, market, mark +
<kn> /n/
knight, knee, know
acknowledge, weakness
<l(l)> /l/
left, bullet, shelf, girl, full
half, calf, calm, walk, could
<m(m)> /m/ morning, comma, room +
<mb> /mb/
emblem, tumble
lamb, climb, tomb
<mn> /m/
damn, hymn
hymnal, gymnasium, omniscient
n(n) /n/ name, manner, mean +
<n[cgkqx]1> /ŋ/
/ŋg/, /ŋk/
singer, long
finger, anxiety, thinker, inquire
<p(p)> /p/ plan, copper, temple, map +
<ph> /f/
/ph/, /p/
photography, physical
<p[nst]> /pn/
<q> /kw/

/ə/ or lengthening of preceding vowel
red, mirror
centre, tender, mirror, core
<s(s)> /s/
sound, master, moss, close
measles, close
sugar, sure
<sc> /sk/
descent, muscle
<sh> /ʃ/
shore, fashion, flash
<sw> /s/
swine, swindle, swat
<t(t)> /t/ top, bitter, hat +
<th> /θ/
theme, both, monthly
the, these, brother
<v> /v/ van, moving, curve +
<w> /w/ water +
<wh> /h/
who, whole
where, when, why, what
<x> /ks/
box, mixture
exact, exist
Xerox, xenophobia, xylophone
<y> /j/ yellow, yoghurt, yoke +
<z(z)> /z/ zone, dazzle, buzz +

  1. Go through the table above and try to find further examples (or counter-examples) for the individual combinations.
  2. Try to understand how each pattern is conditioned by its environment(s) or whether certain cases/exceptions do not follow any rules.
  3. If you’re unsure about certain words, check them in a pronouncing dictionary.

Vowel Graphemes/Grapheme Combinations and Their Realisations

As you will see form the table below, the situation is much less straightforward for the realisations of vowel graphemes or their combinations. This table represents an attempt to illustrate the diversity of possible realisations that exists for the various accents of English as spoken by native speakers and is mainly intended to raise your awareness concerning this issue. It should by no means be seen as exhaustive and will certainly be more accurate in its representation of the potential realisations of British English, although North American, Australian and New Zealand accents are also covered to some extent, without being referred to explicitly. Many of the details occuring in this table will be covered in later sections of the course dealing with individual accents of English, but for now, you should at least be aware of the fact that whenever an [ɑ(ː)] realisation occurs as a potential option, this ought to be interpreted in a somewhat different way when it refers to either a (Southern) British or an American realisation. In the British variant, this ought to be interpreted as having a quality that is distinctly further back than its American ‘counterpart’, which is somewhat nearer to [a(ː)]. I am using the tilde symbol (~) to indicate a range of pronunciations in more or less the same way as John Wells does in his Accents of English.

<a> [a~æ]cat, bass+
[eː~eɪ~ʌɪ]base, bass+
[a~aː~ɑː~ɑ˞~æ(ː)]bath, car+
[ɔː~ɔ˞~oː]water, warm, wall+
[ɒ~ɑ(ː)]watch, what+
[ə~ɐ]about, ago, again+
[ɪ~ə]cottage, luggage+
<a[iy]> [eː~eɪ~ʌɪ]main, stray, bait+
[ɛː~ɜː~ɛə]air, fair+
/ə/captain, fountain+
[ɛ~e]said, says, again+
<a[uw]> [ɑː~ɔː~oː]faun, prawn +
<e> [ɪ~e~ɛ]emerald, bet+
<ea> /eɪ~iː/eat, beat, tea+
<ee> [eɪ~iː] eel, meet, see+
<e[iy]> [eː~eɪ~ʌɪ]reign, fein+
<e[uw]> /juː~uː/ ewe, euphemism, few+
<i> [ɪ~ə]interest, tip+
[aː~aɪ~ɒɪ]five, fire, pirate+
[ɜː~ɜ˞~ɛː~ɜɪ]bird, third+
<ie> [aɪ~ɒɪ]pie, trie+
[aɪə~ɒɪə~aːə]diet, quiet+
<io> /ə/ notion, passion, fashion+
<o> [ɒ~ɑ(ː)]option, pot+
[əʊ~oː~ɒ]omega, over, potato+
[ɔː~oː]tore, torn+
<oa> [ɔː~o:]oar, board+
<oe> [uː~yː]shoe+
<o[iy]> /ɔɪ/ noise, hoist, boy, buoy+
/əʊʷɪ/2 coincidence+
<oo>[uː~yː]ooze, moon, shoo+
[ʊ~yː~ɤ]foot, book+
<o[wu]> [ɔː~oː]four+
[aʊ~oʊ~əʊ]out, shout, cow+
<u> /juː/user, usual+
[ɜː~ɜ˞~ɛː]fur, curl+
<ue> /uː/Sue, queue+
<ui> /uː/fruit, cruise+
/ɪ/build, guild+
[aɪ~ɒɪ]guide, quite+
<uy> [aɪ~ɒɪ] buy, guy+
<y> [aɪ~ɒɪ]cry, fry, try+
/j/yesterday, yearn+

  1. Although we won’t be dealing with different accents until later on in the course, try to got through the above table and attempt to spot which realisations may occur in which country/region. In order to achieve this, try to make use of what we have learnt about the tongue positions and how they are indicated on the vowel chart.
  2. Try to identify regular or semi-regular patterns, paying particular attention to the environment each example occurs in in terms of surrounding consonants, as well as aspects of morphology.

Phonemes & Allophones

So far, we’ve been talking about sounds and phonemes as if they were more or less the same thing. Neither have we yet made any attempt to give a proper definition of the term phoneme, but this is obviously important because the sounds (or phones) we have discussed earlier may in fact represent a variety of different conceptual entities, as the grapheme-phoneme correspondence tables above show. So what actually are phonemes? Probably the simplest explanation is that a phoneme is an abstract concept used to represent a group of sounds or sound combinations that are similar enough to each other to be preceived as performing the same function in a speech chain. A standard illustrative example should clarify this further: as we have already seen, in English, word-initial voiceless plosives – such as in pit, pot, put, tip, top, etc. – are generally considered to be aspirated, i.e. they are produced with a slight hissing sound immediately after the plosive burst and before the onset of the following vowel. This is usually transcribed using a raised h ([ʰ]), or sometimes [s], if there is strong aspiration. But as far as realisations in English accents go, an initial plosive need not always be aspirated, in which case it may be marked with [=] and will sound rather similar to the voiced variant. In terms of the function of the initital plosive, it does not really matter whether we have strong, normal or no aspiration; we will usually understand the word in which it occurs correctly, either because we’re aware of the context in which the plosive is realised, or because we’re used to/aware of the particular – possibly even idiosyncratic – accent of the speaker. So, all that matters in this context is that we can identify the sound we heard as a realisation of an initial plosive. We can see something very similar to this in the case where a voiceless plosive is preceded by a voiceless alveolar fricative /s/, as in the word sports, where the plosive is generally de-aspirated, thus yielding [sp=o:ts], and we could never really be tempted to confuse it with the non-existant sborts.

Only in relatively rare cases do we possibly get confused by features like this, and those tend to be the cases where we actually have words that constitute examples of minimal pairs, i.e. word pairs that are only distinguished by having one sound in them that makes them different from one another. Some examples for minimal pairs would be /pɪt/ & /bɪt/, the ‘classic’ /ʃɪp/ & /ʃiːp/, /kap/ & /kat/, etc.

The idea of the phoneme is mainly based upon the fact that we can establish distinctions of meaning between words by replacing certain elements, i.e. sounds, by one another. One way in which we can distinguish the elements that can replace each other is to use a minimal pair test in the way we have just seen in the examples above. Another way of identifying the inventory of phonemes used in a language is to look at their distribution. Another classic example here is the difference between the occurrences of /h/ & /ŋ/ in English, where the first can never occur at the end of a word – other than in the form of aspiration – and the second never at the beginning. Of course the two units we can distinguish in this way also need to be sufficiently different from one another in the way they are produced, otherwise it would not make any sense. So, returning to our example of the voiceless plosives above, we could say that the absence of voicing and same place and manner of articulation in all examples, including the positional variant after the fricative, makes these sounds sufficiently similar to each other to count them as one phoneme. On the other hand, if we add the voicing to e.g. the bilabial plosive, we do get a distinction in meaning between minimal pairs like /pɪt/ & /bɪt/, so that we can assume that there are two different phonemes.

In cases where we have instances of the same phoneme, but marginally different realisations, we speak of allophonic variation or allophones. This term comes from the Greek word αλλο, which simply means other. Further examples for this are the occurrence of ‘clear’ (/l/) and ‘dark l’ (/ɫ/) in (many, if not most, accents of) English, where the latter only occurs in final position and the difference in the pronunciation of /k/ in the words key and coo, where the obstacle for the plosive in the former is made considerably further to the front than for the latter articulation, due to the nature of the following vowel.

Sources & Further Reading:

Knowles, G. (1987). Patterns of Spoken English: an Introduction to English Phonetics. London: Longman.

Wells, J. (1982). Accents of English (Vols. 1-3). Cambridge: CUP.

Wells, J. (2000). Longman Pronunciation Dictionary (2nd ed.). London: Longman.