The Syllable and Phonotactic Constraints
Jonathan Harrington and Felicity Cox
Phonotactic Constraints
We have seen in the preceding section that all languages build their words from a finite set of phonemic units. It is also true that in all languages there are constraints on the way in which these phonemes can be arranged to form syllables. These constraints are sometimes known as phonotactic or phoneme sequence constraints and they severely limit the number of syllables that would be theoretically possible if phonemes could be combined in an unconstrained way. Some simple examples of phonotactic constraints in English include: all three-consonant clusters at the beginning of a word start with /s/ ('sprint', 'squire', 'stew' etc); nasal consonants cannot occur as the second consonant in word-initial consonant clusters unless the first consonant is /s/ (e.g. there are no words in English than begin with /bm dn/ etc), although this is certainly possible in other languages (e.g. German which allows /kn/ in words like 'Knoten', meaning 'knot' - we can see from the spelling that English used to allow this sequence as well). Another important point about phonotactic constraints is that they vary from language to language, as this example of English and German has just shown.We will consider firstly why languages have phonotactic constraints. The main reason is to do with the limits on the talker's ability to pronounce sequences of sounds as one syllable, and the listener's perception of how many syllables he or she hears from a given sequence of phonemes. Consider for example a sequence like /pʁ/ i.e. a voiceless bilabial followed by a voiced uvular fricative. Most of us with some training can produce this sequence (e.g. /pʁa pʁit/ etc.) as a monosyllabic word even though it doesn't occur in English. Now try reversing the order of the cluster. With some phonetics training, you could almost certainly produce /ʁp/, but what is much harder (even for a trained phonetician) is to produce the sequence before a vowel such that the resulting sequence is monosyllabic. For example, try /ʁpi/ -- even your best attempts at producing the /ʁ/ followed by the /p/ will probably still lead to a percept of two syllables when /ʁp/ precedes a vowel.One of the main reasons, then, why languages have phonotactic constraints is because their sequential arrangement is itself a cue to the number of syllables in a word. When we produce an English word like 'print' for example, we want to convey to the listener not only that this word is composed of a certain number and type of phonemes, but also that the word happens to be monosyllabic: and the listeners' perception of how many syllables there are in a word depends to a certain extent on the arrangement of phonemes in sequence, as we saw from the example of /pʁ/ and /ʁp/ that has just been given.In order to explain why listeners hear e.g. /pʁi/ as one syllable, but /ʁpi/ as two, we need to appeal to what has been called the syllable's sonority profile.
Sonority Profile
Sonority is an acoustic-perceptual term that depends on the ratio of energy in the low to the high part of the spectrum, but it is also closely linked with the extent to which the vocal tract is constricted. In general terms, open vowels like [a] have the highest sonority because the vocal tract is open and a large amount of acoustic energy radiates from the vocal tract. At the other extreme, voiceless oral stops have least sonority because there is no acoustic energy during the closure in which the vocal tract is constricted. Languages prefer to build syllables with the most vowel-like sounds nearer the middle, and the least vowel like sounds (=oral stops, voiceless fricatives) near the edge(s). Syllable structured in this way are said to conform to the sonority profile.
i.e. oral stops are less sonorous than fricatives which are less sonorous than nasals etc. If they conform to the sonority profile, consonants sequences in syllable onsets increase in sonority from left to right and consonant sequences in syllable codas decrease in sonority from left to right. From this we can predict which consonant sequences are more probable for syllable onsets and codas.
Jonathan Harrington and Felicity Cox
Phonotactic Constraints
We have seen in the preceding section that all languages build their words from a finite set of phonemic units. It is also true that in all languages there are constraints on the way in which these phonemes can be arranged to form syllables. These constraints are sometimes known as phonotactic or phoneme sequence constraints and they severely limit the number of syllables that would be theoretically possible if phonemes could be combined in an unconstrained way. Some simple examples of phonotactic constraints in English include: all three-consonant clusters at the beginning of a word start with /s/ ('sprint', 'squire', 'stew' etc); nasal consonants cannot occur as the second consonant in word-initial consonant clusters unless the first consonant is /s/ (e.g. there are no words in English than begin with /bm dn/ etc), although this is certainly possible in other languages (e.g. German which allows /kn/ in words like 'Knoten', meaning 'knot' - we can see from the spelling that English used to allow this sequence as well). Another important point about phonotactic constraints is that they vary from language to language, as this example of English and German has just shown.We will consider firstly why languages have phonotactic constraints. The main reason is to do with the limits on the talker's ability to pronounce sequences of sounds as one syllable, and the listener's perception of how many syllables he or she hears from a given sequence of phonemes. Consider for example a sequence like /pʁ/ i.e. a voiceless bilabial followed by a voiced uvular fricative. Most of us with some training can produce this sequence (e.g. /pʁa pʁit/ etc.) as a monosyllabic word even though it doesn't occur in English. Now try reversing the order of the cluster. With some phonetics training, you could almost certainly produce /ʁp/, but what is much harder (even for a trained phonetician) is to produce the sequence before a vowel such that the resulting sequence is monosyllabic. For example, try /ʁpi/ -- even your best attempts at producing the /ʁ/ followed by the /p/ will probably still lead to a percept of two syllables when /ʁp/ precedes a vowel.One of the main reasons, then, why languages have phonotactic constraints is because their sequential arrangement is itself a cue to the number of syllables in a word. When we produce an English word like 'print' for example, we want to convey to the listener not only that this word is composed of a certain number and type of phonemes, but also that the word happens to be monosyllabic: and the listeners' perception of how many syllables there are in a word depends to a certain extent on the arrangement of phonemes in sequence, as we saw from the example of /pʁ/ and /ʁp/ that has just been given.In order to explain why listeners hear e.g. /pʁi/ as one syllable, but /ʁpi/ as two, we need to appeal to what has been called the syllable's sonority profile.
Sonority Profile
Sonority is an acoustic-perceptual term that depends on the ratio of energy in the low to the high part of the spectrum, but it is also closely linked with the extent to which the vocal tract is constricted. In general terms, open vowels like [a] have the highest sonority because the vocal tract is open and a large amount of acoustic energy radiates from the vocal tract. At the other extreme, voiceless oral stops have least sonority because there is no acoustic energy during the closure in which the vocal tract is constricted. Languages prefer to build syllables with the most vowel-like sounds nearer the middle, and the least vowel like sounds (=oral stops, voiceless fricatives) near the edge(s). Syllable structured in this way are said to conform to the sonority profile.
i.e. oral stops are less sonorous than fricatives which are less sonorous than nasals etc. If they conform to the sonority profile, consonants sequences in syllable onsets increase in sonority from left to right and consonant sequences in syllable codas decrease in sonority from left to right. From this we can predict which consonant sequences are more probable for syllable onsets and codas.
probable /pla fni lju sma pfle/ /alp ims ort/ | less probable /lpa nfi jlu lfpe/ /apl ism otr/ | Why? The syllables on the right have two sonority peaks -- and so it's much more difficult to produce them so that they sound like one syllable…for example: |
So a language is more likely to build monosyllabic words from the combination of phonemes on the left than on the right.
Languages prefer to build syllables from phonemes such that the sonority rises from the left syllable edge, then reaches a peak (at the vowel), and then falls. Therefore, a language is more likely to have a syllable like /pla/ than /lpa/, because in /pla/ the sonority rises from its lowest value for /p/, increasing for /l/, and reaching a peak with /a/. Similarly, a language is more likely to have /amp/ than /apm/. We can now see why listeners might hear two syllables in /ʁpa/ even if a talker intends only one:
because the sonority is higher for /ʁ/ (since it is a fricative), then falls for /p/, then rises again for /a/ (and the condition to hear one syllable would be that there is a progressive rise in sonority from the syllable's left edge).It must be recognised that there is only a tendency for syllables to conform to the sonority profile. So while most syllables do conform to the sonority profile in English, many syllables that contain a consonantal cluster with /s/ do not. An example of a syllable that does conform to the sonority profile is 'flounce', phonemically /flæɔns/ in (Australian) English. In the initial consonant cluster, /f/ is less sonorous than /l/ which is less sonorous than the diphthong; in the final consonant cluster, the diphthong is more sonorous than /n/ which is more sonorous than /s/ and so the sonority rises from the left edge of the syllable, reaches a peak at the diphthong, and then falls over the final cluster. But a word like 'spin' violates the sonority profile (because /s/ is more sonorous than /p/) and so does 'act' (because /k/ and /t/ are equally sonorous). The sonority profile is therefore a general tendency which determines many, but by no means, all phonotactic constraints.
Phonotactic Constraints: Syllable Onset, Coda and RhymeWhen discussing phonotactic constraints, it is helpful to structure the syllable hierarchically in terms of an onset and a rhyme, and sometimes also the syllable coda. See the section on "Syllable Structure" for more details.
We can then discuss phonotactic constraints: within the onset
within the coda
within the rhyme
The most extreme phonotactic constraints (extreme in terms of the greatest restrictions in the sequential arrangement of phonemes) are in the onset. For example, in English: /f/ can only be followed by approximants (as in 'fly'), there are no consonant phonemes that can follow affricates etc.The phonotactic restrictions in the coda in English are often (but not always) a mirror-image of those in the onset (as you'd expect if the syllable's legal phoneme sequences are strongly influenced by the sonority profile). For example, English allows /pl/ in the onset ('play') and /lp/ in the coda ('help'); it allows /fr/ in the onset ('free') and, for rhotic dialects (e.g. Gen. American English), /rf/ in the coda ('surf'). But there are also many permissible coda sequences that are allowed whose mirror-image is disallowed in the onset (e.g. /mp/ as in 'lamp', but no /pm/ in the onset).
Finally, there are far fewer restrictions in the rhyme -- these are to do with the restrictions on nucleus-coda combinations. But as an example of a rhyme constraint, there are no long vowel + /ŋ/ sequences (no words like 'seeng', 'flowng', although the onomatopoeic 'boing!' is allowed).
Language-specific constraintsLanguages differ in the kinds of onsets they allow:
/kn/ | /skw/ | /sb/ | /vr/ | |
English | no | yes | no | no |
German | yes | no | no | no |
French | no | no | no | yes |
Italian | no | no | yes | no |
In English the maximum number of consonants that can make up the syllabic onset at the beginning of an isolated word is three. The first can only be /s/, the second has to be /p, t, k/, and the third has to be an approximant /w, j, r, l/.
eg. | splayed | strayed | scrape |
| spew | stewed | skewed |
| squish | squawk | squeal |
These are all CCCVC When the third consonant is /w/ then the first two must be /sk/Whilst /spr/ and /str/ are permitted syllable-initially, /spw/ and /stw/ are not permitted syllable-initially in English.Most languages do not allow as many as three consonants in the syllabic onset however there are some that allow up to six.Restrictions in the coda are often the mirror image of those in the onset, eg pl ~ lp due to the sonority principle. However there are many exceptions eg /nd/ in "end" but not /dn/.
The number of final consonants in an English rhyme can range from one to four. eg. /sɪk/ sick, /sɪks/ six, /siksθ/ sixth, /siksθs/ sixths
Languages differ in the structures that they permit. English permits complex codas and onsets. Languages like Hawaiian, for instance, only allow a single consonant in the onset and none in the coda, so every syllable ends in a vowel. Standard Chinese allows only nasal consonants in the coda, so syllables are either open or closed with a nasal.
Phonotactic constraints: Combinatory and DistributionalLanguages differ in the structures that they permit. English permits complex codas and onsets. Languages like Hawaiian, for instance, only allow a single consonant in the onset and none in the coda, so every syllable ends in a vowel. Standard Chinese allows only nasal consonants in the coda, so syllables are either open or closed with a nasal.
Some Combinatory Constraints in English
- /ŋ/ cannot be preceded by long vowels or diphthongs
- /tʃ, dʒ, ð, z/ do not cluster
- /r, w, l/ only occur alone or as non initial elements in clusters
- /r, h, w, j/ do not occur in final position in Australian English, but /r/ can occur in final position in rhotic dialects such as American English.
- In final position only /l/ can occur before non-syllabic /m/ and/n/.
- /ŋ/ cannot occur word initially
- /e, æ, ɐ, ʊ, ɔ/ cannot occur word finally
- /ʊ/ cannot occur initially
- /ʒ/ only occurs initially before /ɪ, iː, æ, ɔ/ in foreign words such as genre.