Laying the foundations: sound systems in language

While the phonetician studies the articulatory, auditory or acoustic properties of speech sounds, the phonologist’s task is to understand how these sounds are organized in languages. In any language, a relatively small number of contrastive speech units, known as phonemes, are responsible for distinguishing all real and potential words, and native speakers quickly learn which distinctions matter for this purpose and which do not; they are also aware of the rules by which these phonemes may or may not be combined. In this chapter, we will look in detail at the concept of the phoneme, and consider some tests by which they may be identified.

Two speakers of the same language may have slightly different phoneme inventories, or they may use the same set of phonemes in different ways. We will learn how we as speakers recognize these accent differences, and how phonologists attempt to model what speakers unconsciously ‘know’ about the sound system of their mother tongue. We will also consider syllable structure, which proves a very important factor in determining how the sounds of a language are organized.

Phonemes and allophones

If two native English speakers were asked to say the sequence [bat] ten times, it is likely that none of their [b]s, [a]s or [t]s would sound exactly the same, but a third English speaker would nonetheless understand the same word, bat, every time. This is because speakers quickly learn to distinguish the differences that matter in their language from those that do not. An English-speaking child will very soon learn when a [p] sound, if voiced, may become a [b] and that this difference is important because pull and bull, pat and bat, path and bath and so on mean different things. And if the voicing distinction is important for [p]/[b], then it is likely also to be important for [k]/[g], [t]/[d] and so on.

The child will soon realize, however, that other differences are functionally unimportant in this sense. For most British English speakers, for example, the sounds represented orthographically by l at the end of cool and the beginning of leap are quite different, but speakers think of them as ‘the same’ sound. Many Cockney or Glaswegian speakers use a glottal stop \[

] intervocalically in words like water or matter and yet the words will be perfectly well understood as if the speaker had produced a [t] (indeed, even RP speakers often use glottal stops in words like Gatwick or fortnight, where they generally pass unnoticed). Speakers home in, then, on the distinctions (or oppositions as they are known) which are crucial for doing what speech sounds ultimately need to do in language, i.e. distinguish words, and ignore those that do not.

Case study: Lewis Carroll and the art of English phonotactics

As we saw in Chapter 4, languages demonstrate remarkable economy in using a small number of resources to produce all the words they need, not just for their existing lexicon (or stock of words), but also for an ever-growing number of new words: quidditch, satnav, website and selfie were all unknown 30 years ago, but have settled comfortably into the English language. This creativity is not haphazard but rule-governed: English speakers would not similarly have accepted new words like bzork, thlick or drailx, even though they use familiar English sounds. The nonsense words of Lewis Carroll’s famous poem Jabberwocky, on the other hand, pose no problem:

’Twas brillig, and the slithy toves

Did gyre and gimble in the wabe:

All mimsy were the borogoves,

And the mome raths outgrabe.

No native English speaker – probably not even Carroll himself – can say what brillig, gyre, wabe or borogoves mean (though some online have had a stab at doing so), but the poem works because all these words, and the many other unfamiliar inventions elsewhere in Jabberwocky, could exist in English: they just happen not to. They represent what linguists call accidental gaps in the lexicon. Carroll’s creations all respect the phonotactics of English, i.e. the constraints on the way its speech sounds can be combined.

Phonotactics are highly language-specific. While English allows words to begin sp-, st- or sk-, for example, Spanish does not, and Spanish learners of English often struggle with these forms initially (‘I espeak Espanish’, etc.). English, by contrast, rules out initial sequences such as vzd- or vn-, which present no difficulty to Russian speakers in such words as vzdor (‘nonsense’) or vnuk (‘grandson’).

The distinctive speech units of a language are known as phonemes, and these provide the essential building blocks from which all well-formed words (or lexemes) in that language are produced. Phonemes can therefore be thought of as the atoms of a language, and just as atoms have subatomic particles, so phonemes divide into smaller units known as allophones.

Let’s look again at the leap and cool examples above. For most (but not all) British English speakers, the two sounds are ‘clear’ l [l] in the former and ‘dark’ l [ɫ] in the latter. These sounds do not have the potential to distinguish words: pronouncing cool with ‘clear’ l results in a pronunciation which might correspond to that of Irish or Welsh English speakers (who generally use clear l in all positions) but the same word, cool, will be understood. Using ‘dark’ l in leap likewise might make your pronunciation sound slightly Russian (Russian does not have clear l), but the meaning would be unchanged. If, by contrast, we were to replace the [l] of leap with a [h], a [k] or a [p] sound, then different words (heap, keep and peep) would be understood.

Clearly there’s something more important from an English speaker’s point of view about swapping \[l] for \[h] than \[l] for [ɫ] in this environment. A phonologist would say simply that \[l] and \[ɫ] are allophones of the phoneme /l/, which English speakers perceive to be in some sense ‘the same’. It is important to realize that languages can organize the same sounds in different ways: in Polish, for example, the distinction between clear and dark l is phonemic, i.e. they do have the potential to distinguish words, as for example in luk ‘skylight’ and łuk ‘bow’.

…our native phoneme system tends to get in the way when we try to learn other languages. It is perhaps unsurprising that we should find it difficult at first to produce sounds which do not figure at all in our first language. However, it is just as difficult, and sometimes worse, to learn sounds which are phonemically contrastive in the language we are learning, but allophones of a single phoneme in our native system. For instance, there is no contrast between aspirated \[`t`^h^] and unaspirated `[t]` in English: we can predict that the former appears only word-initially. In Chengtu Chinese, however, /`t`/ contrasts with /`t`^h^/, as we find minimal pairs like `[tou]` ‘a unit of dry measure for grain’ versus \[`t`^h^`ou`] ‘to tremble’.(McMahon 2002: 20)

Two points need to be made here. Firstly, you may have noticed a subtle notational change in the previous paragraph. When we referred to ‘the phoneme /l/’, the square brackets we have consistently used for speech sounds were replaced by slants. This is because, when we talk of a phoneme, we refer not to a speech sound but to an abstract unit. The phoneme /l/ might be pronounced or realized[l] or [ɫ], according to the context, and since the phoneme /l/ doesn’t necessarily mean the sound [l], we could in principle use any symbol we liked between the slants, for example /

/, /

/ or /

/. In practice, however, linguists generally use an IPA symbol which corresponds to one of the more common allophones. This can occasionally cause confusion, as in the case, for example, of the phoneme /æ/ in words such as hat, pack and map. Here /æ/ was chosen because it corresponded to the (now rather old-fashioned) RP pronunciation that was de rigueur for BBC newsreaders until around the 1960s. These days, of course, most English speakers say [hat] rather than [hæt], but textbooks retain /æ/ by convention to indicate the vowel which distinguishes hat from hot, hit, hut, height, etc., even though the phoneme is realized in a variety of ways and relatively rarely as [æ].

Key idea: The minimal pair test

The distinctive speech units of a language are called phonemes: phonemes may have several realizations, or allophones. Phonemes and allophones can be determined by the minimal pair test: if commuting two sounds in a word results in a change of meaning, then the sounds are in phonemic opposition; if they do not, the sounds are allophones of the same phoneme.

Secondly, in those varieties that use both allophones it is possible to predict the range of environments or distribution of each. Clear l \[l] occurs prevocalically at the start of a syllable, while dark l \[ɫ] occurs in postvocalic positions. Because the two sounds cannot occur in the same environment for these speakers, they cannot contrast words. As more than one linguist has put it, they are a bit like Superman and Clark Kent: never seen together in the same place, because each is one part of the same whole. This is known as complementary distribution.

To take another example, try saying copy and kitchen, paying particular attention to the two initial consonants. For most English speakers, it’s highly likely that your tongue will leap forward for the second word, and although they sound similar, it should be evident that the two consonants are in fact distinct: the first is a velar plosive, pronounced at the soft palate at the rear of the vocal tract, while for the latter the sound is produced further forward, at the hard palate behind the alveolar ridge. Using the terminology we learned in Chapter 4, the sounds are therefore voiceless velar and palatal plosives respectively, with the IPA symbols [k] and [c] (the wrong way round from the point of view of the spelling!): their voiced equivalents, heard in got and give respectively, have the symbols [g] and \[

]. The distribution is therefore phonetically conditioned, in that the fronted consonants [c] and \[

] occur before front vowels such as [i] while [k] and [g] occur before back vowels such as [ɒ].

One can see how this might make life easier for the English speaker: by advancing the tongue in readiness for a front vowel and retracting it for a back one, a little articulatory effort is saved. But not all languages work in the same way: in Hungarian, for example, the palatal/velar contrast is phonemic, rather than allophonic as in English. To determine whether two sounds are phonemes, phonologists apply the minimal pair test: can we identify two different words that differ only by virtue of one sound? On this test, /k/ and /t/ are phonemes of English on the strength of kick/tick, pick/pit, school/stool and many other pairs. Clear and dark l on the other hand have no pairs in which they contrast, because they never occur in the same environments (and, as we saw, if we force the issue by switching them in their respective environments, we do not get a change in meaning), so these sounds are both allophones of /l/.

Key idea: Complementary distribution

Where allophones are in complementary distribution, they occur in different environments and are therefore unable to be contrastive.

For a final example of complementary distribution, you’ll need a single sheet of paper. Hold it about 3–4 cm in front of your mouth and say pip several times, and then bib; now do the same thing with kick and gig. You will notice that the paper moves considerably for pip and kick but less so for bib and gig, and (since the vowel is unchanged) you might suspect that this has something to do with the consonants involved. What matters here is whether the first consonant is a voiceless plosive like /p/ or /k/: these are pronounced with greater articulatory force than their voiced counterparts – hence the alternative terms fortis and lenis (‘strong’ and ‘weak’) for voiceless and voiced consonants respectively – and at the start of stressed syllables are accompanied by a small outbreath or aspiration, represented in narrow transcription by a superscript \[^h].

Now try the paper experiment again, this time with pin and spin: this time the paper moves in the former but not the latter case. This is because aspiration of /p/, /t/ and /k/ does not happen after /s/, so pin is realized \[p^hIn] but spin is [spin]. Again it is an arbitrary fact about English that aspiration is non-phonemic, i.e. cannot be used to distinguish words, as it can in Thai:

• [paa] ‘forest

• \[p

• aa] ‘to pound’

• [tam] ‘to split’

• \[t

• am] ‘to do’

In some cases, the differences between allophones of a given phoneme may be quite wide. The distance in articulatory terms, for example, between the closure of the glottis required for a glottal stop \[

] and the closure formed between tongue and teeth or alveolar ridge for [t] could hardly be greater, but Cockney, Glaswegian and many other English speakers perceive them as being ‘the same sound’ in that one may say water as [wɔ:tƏ] or [wɔ:Ə] and mean the same thing (see Case study below).

Case study: Free and not-so-free variation

As we saw, [t], \[t^h] and \[

] are all allophones of /t/. There is, however, an important difference here between the distribution of [t]and \[t^h] on the one hand, which are in complementary distribution and cannot occur in the same environments, and \[

] on the other, which appears (at least in some environments) to substitute freely for [t], as in the water example above: many speakers in fact switch between the two. Similarly, in Spanish, which does not have a phonemic opposition between /b/ and /v/, the word Vale! (‘Ok!’) can be pronounced [bale], [vale] or [βale]([β] is the IPA symbol for a voiced bilabial fricative), and be understood in the same way in each case. Alternation of this kind, which is unconstrained by phonetic environment, is known as free variation and was long dismissed by linguists as being of little theoretical interest – a case of ‘You say tomayto, and I say tomahto’, if you like – and scarcely worthy of comment (‘…so let’s call the whole thing off’). There is, however, a very important difference between water as [wɔ : tƏ] and as [wɔ :Ə], namely that the latter pronunciation, in British English at least, is of considerably lower status than the former, something of which most speakers are acutely aware. The variation, then, is socially rather than phonetically conditioned.

Linguists long chose to ignore this obvious fact on the grounds that social data should not be allowed to intrude upon linguistics if it were to establish itself as an autonomous discipline and be taken seriously as a science (see Chapter 2). This consensus was challenged in the 1960s by sociolinguists who argued that no satisfactory explanation of language change could be provided without taking account of social factors. As we shall see in Chapter 11, their findings have forced us to revise our view of how ‘free’ so-called free variation actually is.

Instead of calling this ‘free’ or ‘sporadic’ variation, and abandoning the field, we will pursue the matter further, using every available clue to discover the pattern which governs the distribution. (Labov 1972: 9)

Spotlight: Friends, aspirate stops and ‘stage spitting’

Aspiration of voiceless stops is not merely an accidental fact of English pronunciation, it’s a very important cue which helps English speakers to distinguish /p/ and /b/, /k/ and /g/ and /t/ and /d/. In the theatre, normal aspiration may be difficult for the audience to hear, and needs to be projected.

In the US sitcom Friends, series 7 episode 23 (see http://www.youtube.com/watch?v=0ISJS4gSBh0), inexperienced actor Joey Tribbiani (Matt Le Blanc) and his illustrious co-star Gary Oldman nearly come to blows over the latter’s spitting on stage. This is only averted when Oldman explains that this is the theatre trick of ‘enunciation by spitting’, which exaggerates the natural aspiration of the voiceless stops /p/, /t/ and /k/ for an audience who may not otherwise be able to hear it from the stage.

The phoneme: problems and solutions

We have seen a number of examples of complementary distribution, in which allophones of the same phoneme occur in different environments and therefore lack the potential for functional contrast. In some cases, however, we can identify sounds whose distribution is certainly complementary, but which, for other reasons, we would not wish to consider as members of the same phoneme. The most celebrated example of this kind is that of [h] and [ŋ] (generally represented by the digraph ng in conventional orthography), which have very different, and certainly complementary, distributions:

• hear   ring

• hope   singing

• Henry   singer

• ahead   long

• ahoy!   alongside

While [h] only occurs syllable-initially, [ŋ] is only to be found syllable-finally. These sounds seem to meet the criteria for conditioned allophones, and there are no minimal pairs like hope/\*ngope or ring/\*rih, so one might want to suggest that they are members of a single phoneme (which we can call ‘heng’ for convenience).

In fact, there are good reasons for rejecting ‘heng’. Firstly, the two sounds, a glottal fricative and a velar nasal, seem very different in kind: in other words, they fail the test of phonetic similarity. But as our /t/ example showed earlier, allophones can be very dissimilar, so it would be dangerous to give undue weight to this criterion alone. A more important consideration is that there is a natural class of consonants, nasals, to which one member belongs but the other does not, and that this grouping shares a number of properties which [h] does not. Like /n/ and /m/ (but unlike [h]), [ŋ] has fortis and lenis oral equivalents, and although its distribution is restricted when compared to /n/ and /m/ (both of which can occur syllable-initially or syllable-finally in English), it behaves exactly like the other nasals, notably with regard to the homorganicity rule for nasal+oral stop sequences, as we will see later.

None of these things are true for the glottal fricative [h], and no meaningful generalizations about [ŋ] can be framed which could include [h]. It therefore seems intuitively and practically more sensible to view these sounds as separate phonemes /h/ and /ŋ/, with restricted distributions in English. The ‘heng’ question does raise an important problem in linguistics, however: where two competing explanations of the same data are available, how should one choose between them? Generally the principle of Occam’s Razor is applied in such cases, namely that descriptively economical theories should be favoured until simplicity can be traded for greater explanatory power. Positing a ‘heng’ phoneme appears descriptively elegant in that it captures a distributional fact about the two sounds involved, but it fails to capture a range of other properties which one putative ‘allophone’ shares with another group of English phonemes.

An assumption we have made thus far, but not stated directly, is that allophones must belong only to a single phoneme: this is known as the biuniqueness condition. It is not difficult to see that if allophones could belong to several phonemes, the ensuing ambiguities would make language much more difficult to process. But the important working principle that allophones belong to one and only one phoneme encounters some notable problems. Consider the following examples from German:

• Das Rad [rat] ‘wheel’

• Der Rat [rat] ‘council’

Both words are pronounced with final [t] sound, but the difference in spelling here has a rational basis rather than being simply a historical quirk, as is evident from their plural forms: in the plural of Rad (Räder) the \[d] sound is restored, while in that of Rat (Räte) the final \[t] remains. This is not merely a fact about Rad and Rat: a whole range of singular/plural pairs have a voiceless word-final consonant which becomes voiced when a suffix is added; the same pattern is observed too for some imperative and infinitive pairs:

• Bad [bat] – Bäder [bedƏ] bath/s

• Lied [lit] – Lieder [lidƏ] song/s

• Dieb [dip] – Diebe [dibƏ] thief/ves

• Tag [tak] – Tage [tagə] day/s

• Red! [ret] – reden `` talk!/to talk

• Gib! \[gip] – geben `` give!/to give

• Sag! \[zak] – sagen `` say!/to say

• Reib! \[raip] – reiben `` rub!/to rub

The generalization to be drawn here is that voiced consonants are not allowed (or licensed) word-finally in German, but this descriptive statement fails to capture the fact that the \[t]s of Rad and Rat are fundamentally different: one reverts to \[d] in non-final positions, the other does not. We might want to suggest that the final segment of Rad is ‘underlyingly’ /d/, but to do so we have to sacrifice the biuniqueness condition and claim, in effect, that \[t] is an allophone of both /d/ and /t/.

The solution proposed by Nikolai Trubetzkoy of the Prague School was to suggest that, in certain environments, some phonemic oppositions are unavailable or neutralized: this would be the case for the word-final voiced/voiceless contrast in German, which remains available in other positions. A similar analysis can be proposed in the case of nasal+oral stop sequences in English.

Try saying the following words:

• indeed

• input

• increase

• invade

The orthographic n here is deceptive. In the case of indeed, it corresponds to \[n], but the sound normally produced in input is in fact \[m], and in increase it is \[ŋ] (we will come back to invade in a moment). To summarize, the sequences of nasal and oral stops is as follows:

• n + d

• m + p

• ŋ + k

In all of these cases, the nasal consonant and the following oral stop share the same place of articulation: dental-alveolar, bilabial and velar respectively: the nasal consonant, in other words, is homorganic with the following oral stop. Because the nature of the nasal consonant is entirely predictable from the following consonant, the phonemic contrast between /n/, /m/ and /ŋ/ is neutralized in this environment, which we can indicate via a nasal archiphoneme N in this position (archiphonemes are conventionally represented by capital letters). This phenomenon explains the restricted distribution of /ŋ/ which we discussed above. /ŋ/ occurs in sequences represented orthographically by ng, representing syllable-final N+g sequences, where our homorganicity rule would predict that N is realized \[ŋ] before the velar consonant. In most varieties of English, final /g/ in this environment was lost, leaving /ŋ/ in syllable-final position, but not available in other positions (in some varieties this change did not happen: in north-west England, for example, thing is still realized \[θIŋg]). But what of the nasal consonant in invade? /v/ is a labio-dental consonant and our rule predicts, correctly, that the nasal will accordingly be the labio-dental \[ɱ]. \[ɱ] is not a phoneme of English because it occurs only in this position, where no contrasts with other nasals are possible.

Key idea: The biuniqueness condition

The biuniqueness condition states that allophones must belong to one and only one phoneme. In some environments, however, a phonemic contrast or opposition may be unavailable, or neutralized. Phonologists may refer in such cases to archiphonemes, conventionally indicated by a capital letter.

Comparing accents

When a person is described as ‘having an accent’ the phrase usually has prescriptive connotations: the implication is that their speech deviates from standardized pronunciation. In fact, everyone ‘has an accent’, but what does this mean in phonological terms? Essentially, differences between accents amount to differences in phonological systems, which can be of three principal kinds:

• inventory differences, involving the presence of a phonemic opposition in one accent, but not the other

• distributional differences, in which the accents have the same phonemes but use them differently

• realizational (or phonetic) differences, in which the phonemic contrasts available in both accents are identical, but the phonemes are realized differently.

INVENTORY DIFFERENCES

The stock of phonemes a speaker uses is known as his/her phoneme inventory. As we saw in Chapter 4, an RP speaker has around 21 vowel and 24 consonant phonemes, but the precise number may vary: not all RP speakers use all the centring diphthongs, for example. An important difference in phoneme inventory distinguishes northern and southern varieties of English English: because of a historical vowel split which happened in the south, southern varieties have an additional phoneme /ʌ/ which northern varieties lack. This means that southern speakers have a /ʌ/ – /ɷ/ phonemic contrast for pairs such as cud/could, putt/put, which are homophonous for northerners (\[kɷd] and \[pɷt] respectively). Scottish English varieties have an additional consonant phoneme /x/ which English English varieties lack, meaning that loch \[lɒx] and lock \[lɒk] are generally distinguished north of the border, but not south of it (\[lɒk] in both cases).

DISTRIBUTIONAL DIFFERENCES

Two varieties may share the same phonemes, but distribute them differently. Again, the north-south linguistic divide in England provides a good example. Both northern and southern varieties have the /ɑ:/ and /a/ phonemes in their inventories, but they do not always agree on where they are used:

Table 5.1: Distribution of /ɑ:/ and /a/ in southern and northern English varieties of English

Spotlight: Bristol l

Intrusive r occurs in part because loan words ending in unstressed final -a (e.g. India, banana) had no counterparts in English when they entered the language. Speakers therefore assimilated them to something similar, in most cases the final -er lexical set (e.g. reader, carer), for which /r/ is deleted in non-rhotic areas except before a vowel. In Bristol, a rhotic city in the west of England, these words were assimilated not to the -er set but to the -le set (e.g. little, apple), with the result that banana is pronounced ‘bananal’ and pairs such as idea/ideal and area/aerial are homophonous.

Because Bristol is so out of step with other English regions in this regard, Bristol l has become widely ridiculed outside the city and is increasingly avoided by all but lowest-status Bristolians as a consequence. An accent feature that is so salient as to attract mockery is called a stereotype (see Chapter 11).

In similar vein, an appeal to underlying representations can illuminate some otherwise puzzling phenomena in Standard French, where many words have an orthographical final e which is not pronounced (‘mute e’):

• vase \[vaz]

• belle \[mεtʁ]

• maître \[bεl]

• sauce \[SOS]

Similarly, many words end in orthographical consonants that are not realized: trop, les, bon. As we saw in Chapter 1, it is not uncommon for pronunciation and spelling to be out of step, but in this case standard French orthography may reflect some important facts about French phonology. In spite of the fact that ‘mute e’ is not pronounced, many phonologists would suggest that vase is underlyingly /vazə/, and trop underlyingly /tʁop/, with a deletion rule being applied to delete the final segment in each case.

Once more, a seemingly counter-intuitive approach captures a number of important generalizations. Firstly, we know that the final mute -e is an orthographic holdover from a time when word-final schwa was pronounced: \[vazə] was once the normal pronunciation, and remains so in many parts of southern France today. So, rather like our earlier /r/ example from English, our underlying representation captures a synchronic fact about French and a diachronic one. Secondly, even for speakers who don’t normally pronounce the final schwa, there are occasions when this consonant reappears, typically to break up sequences of three consonants, so Jacques est le maître \[mεtʁ] but Maître Jacques \[mεtʁəʒak]. Similarly, many words have latent word-final consonants which appear before a vowel in what is known as liaison, thus les gens \[leʒā] ‘people’ but les amis ‘friends’ \[lezami]. But why is the final consonant deleted in some cases, but not in others? One answer would be that the final \[z] of vase is always pronounced because in its abstract representation it is non-final, and the underlying final schwa ‘blocks’ the deletion rule. The abstraction is arguably therefore a price worth paying for a richer analysis of what French speakers ‘know’ about their language, namely that some word-final consonants are always pronounced, but others are only pronounced in certain environments.

DISTINCTIVE FEATURES AND NATURAL CLASSES

In some of the examples above, we have seen that certain groups of sounds have similar properties, or tend to behave in similar ways. One set of consonants was affected, for example, by a rule involving word-final position in German, which we could identify by a series of rules:

• /b/ → /p/ /\_ ##

• /d/ → /t/ /\_ ##

• /v/ → /f/ /\_ ##

These rules should be read as ‘/b/ becomes /p/ in the environment of a word boundary’ and so on, and we could in principle set out similar rules for all the consonants affected by the devoicing rule, but such an approach misses an obvious generalization about all these consonant pairs, namely that those to the left of the arrow are all voiced and those to the right are all voiceless. We could therefore restate the rule rather more economically in terms of a single distinctive feature ±\[voice].

• C

• → C

• /\_ ##

The sounds affected by this rule form part of the natural class of voiced consonants. We can define a natural class as a group of speech sounds which share the same specification for one or more features. Natural classes often share properties cross-linguistically. Dutch, Russian and Polish, for example, all have a word-final consonant devoicing rule, as do some French dialects. Another natural class is nasals, the feature \[+nasal] being shared by /m, n, ɱ, ɲ, ŋ, 

, 

/, among others. We can extend the principle further and establish feature matrices for phonemes, as for example in Table 5.3 for the velar consonants of Scots English. The feature ±\[continuant] distinguishes oral and nasal stops \[-continuant], for which the airflow in the vocal tract is blocked, from all other sounds, which allow continuous airflow and are therefore \[+continuant]:

Table 5.3: Feature matrix for velar phonemes in Scots English

The four phonemes form part of a single class by virtue of sharing the feature specification \[+velar]. Note that the specification \[+voice] is redundant for /ŋ/ as all nasals are voiced in English, but this feature is distinctive for /k/ and /g/. The feature ± \[aspirated] is not included in our matrix because aspiration is not distinctive in English: its presence or absence in certain environments for voiceless plosives is entirely predictable and not a property of the phoneme itself, as it would be in Thai, where aspiration is phonemic.

Attempts have been made to establish a universal feature set from which the phonemes of all the world’s languages could be defined. A first set was proposed in 1952 by Jakobson, Fant and Halle and was refined for the description of English by Chomsky and Halle in 1968. This set has since undergone a number of changes, with putatively universal features such as ±\[diffuse] or ±\[compact] having been largely abandoned. Some have criticized the emphasis in most feature systems on binarity: it is not entirely clear that ± features are appropriate for vowels, or that they are well equipped to deal with phonemic tone in languages such as Hokkien or Vietnamese. But there is no doubt that features have a role in identifying natural classes, and in capturing important cross-linguistic generalizations.

Key idea: Identifying natural classes

Phonemes can be broken down into their distinctive features. Phonemes that share the same specification for one or more features are said to belong to a natural class.

Syllable structure

We have already seen evidence that syllable structure has important phonological consequences. This was notably the case for the liquid approximants /l/ and /r/ in English: allophonic variation for /l/ depends on position in the syllable, while /r/ is often deleted syllable-finally, but never syllable-initially. Cross-linguistic evidence suggests that syllables (conventionally represented by a lower case Greek sigma σ) have a hierarchical structure consisting of an onset and a rhyme:

Figure 5.1: Syllable structure

Of these elements, only the nucleus is obligatory, and it must be filled either by a vowel or by a sound sufficiently sonorous to occupy the nuclear or head position. This would include, for example, the liquid /l/ in some English unstressed syllables (e.g. in the second syllable in bott**le**) or indeed /r/ or /l/ in some Czech stressed ones (e.g. p_r_st ‘finger’; v_l_k ‘wolf’). The nucleus together with the coda form the rhyme: a phonological definition of the verb ‘to rhyme’ as conventionally used might therefore be ‘to have an identical nucleus and coda’, e.g. bend/trend, ash/smash, cold/bold, etc.

On either side of the nucleus, onset and coda position are occupied by less sonorous items on the sonority hierarchy (see Chapter 4) than the nucleus, so vowels are not permitted but approximants and other consonants can occur in these positions. I, oh and a are examples of English monosyllabic words which consist entirely of a nucleus; he, you and three are examples of syllables with onset and nucleus, but no coda; and finally ill, elf, asked all have nucleus and coda within the rhyme, but no onset.

Key idea: Syllables

Syllables consist of an onset and a rhyme. The rhyme is made up of an obligatory nucleus and a coda. Some languages have obligatory onsets, or rule out codas, but no language has obligatory codas.

Languages constrain this basic structure in different ways: English allows up to three consonants in onset position, but is strict about the three-consonant combinations it permits: the first must be /s/, the second a stop, and the third a liquid (e.g. sprint, scroll, splash); Russian, by contrast, allows much greater flexibility, e.g. /vdrug/ ‘suddenly’; /sxvatit/ ‘seizes’. While there are languages which require onsets (e.g. Arabic) or rule out codas (e.g. Hawaiian), there are none that require codas, supporting the view that the basic syllable structure cross-linguistically is Consonant+Vowel (CV).

Further evidence for the primacy of the CV structure comes from the Maximum Onset Principle, which states that, whenever consonants can be assigned to either onset or coda position, as many as possible should be assigned to the onset, subject to the phonotactic constraints of the given language. So, for example, in the case of a word like express, the proper syllable division is /εk/ + /sprεs/, because */ksp/ is not an acceptable syllable-initial cluster in English, while /sprεss/ is well formed and meets the requirement that the maximum number of consonants be assigned to the onset.

As April McMahon (2002: 110–12) points out, this principle explains a number of otherwise puzzling phenomena. Why, for example, do non-rhotic speakers pronounce the /r/ in carry, even though it follows a vowel, but not in orchard? The answer is that, by the Maximum Onset Principle, the syllable divisions are /ka+ri/ and /ɔːr+təd/ respectively: in the latter case English rules out */rt

/ as an initial syllable sequence, so the /r/ is assigned to coda position in the previous syllable, where non-rhotic speakers delete it (rhotic speakers, of course, maintain /r/ in both onset and coda positions). Likewise, it might seem surprising that wool for most British English speakers has a dark /l/\[

], but its derivative woolly has a clear one \[l]. Again, the Maximum Onset Principle assigns the /l/ in woolly to onset position /wɷ+li/ and thereby predicts, correctly, that clear \[l] will be selected.

Fact-check

1. [1](answers.mdx#rfn5-1)  What characterizes aspiration?

1. It is non-distinctive in English, but distinctive in Thai

1. It is a distinctive feature of Thai and English voiceless stops

1. It is a non-distinctive feature of voiced stops in English

1. It is only present when a stop consonant is preceded by /s/

1. [2](answers.mdx#rfn5-2)  What is the pronunciation of month, cup, bug in northern and southern English accents an example of?

1. A distributional difference

1. An inventory difference

1. A realizational difference

1. Southerners being difficult

1. [3](answers.mdx#rfn5-3)  What sound is represented by n in inferior?

1. \[n]

1. \[ɱ]

1. \[m]

1. \[ŋ]

1. [4](answers.mdx#rfn5-4)  What must all syllables have?

1. An onset

1. A coda

1. A nucleus

1. None of the above

1. [5](answers.mdx#rfn5-5)  What does neutralization mean?

1. That a sound change has left a silent letter in the written language

1. That a sound is deleted in certain environments

1. That a choice of phonemes is always available

1. That a phonemic contrast is not available in certain environments

1. [6](answers.mdx#rfn5-6)  Which of the following is not a continuant?

1. \[p]

1. \[ɹ]

1. \[ɷ]

1. \[f]

1. [7](answers.mdx#rfn5-7)  What is ‘Bristol l’?

1. A perfectly normal assimilation process involving loan words ending in -a

1. A stereotype

1. A distributional difference with respect to other English accents, involving a small lexical set

1. All of the above

1. [8](answers.mdx#rfn5-8)  In English, what onsets do syllables allow?

1. A maximum of two consonants

1. A single vowel

1. /s/+stop+liquid sequences

1. /s/+stop+fricative sequences

1. [9](answers.mdx#rfn5-9)  Which is the correct syllable division for starstruck?

1. /stαː+strΛk/

1. /stαːs+trΛk/

1. /stαːst+rΛk/

1. /stαːstr+Λk/

1. [10](answers.mdx#rfn5-10)  What does the biuniqueness condition state?

1. That all languages have a unique phoneme inventory

1. That allophones must be assigned to a single phoneme

1. That the same allophone can be assigned to more than one phoneme

1. That phonemes may be realized in two different ways

Dig deeper

B. Hayes, Introductory Phonology (Blackwell, 2009), esp. Chapters 2, 3, 4 & 6

F. Katamba, An Introduction to Phonology (Longman, 1989), Chapters 2–5

R. Lass, Phonology (Cambridge University Press, 1984), Chapters 1–3

A. McMahon, Introduction to English Phonology (Edinburgh University Press, 2002), esp. Chapters 4, 5, 8 & 9

I. Roca & W. Johnson, Course in Phonology (Blackwell, 1999), Chapters 2, 4 & 6

Online sources

Wikipedia articles on the phoneme (http://en.wikipedia.org/wiki/Phoneme) and the minimal pair (http://en.wikipedia.org/wiki/Minimal\_pair)