229
Brian Mott*1   UDK 811.111'342.7
University of Barcelona  DOI: 10.4312/linguistica.57.1.229-241
APPROACHES TO THE SYLLABLE: AN ASSESSMENT
1 INTRODUCTION
Since the emergence of Prague School linguistics, the syllable has generally been 
recognized as a fundamental unit in phonological analysis among all the major schools 
of thought in the field. To Kohler’s scepticism (1966) about the universality of the 
syllable in phonology, for instance, Fudge (1999: 370–371) responds: “I would like 
to state my firm belief that the syllable is a phonological universal.” He also insists on 
the fact that the syllable must be defined in terms of its relationship to other linguistic 
constituents: “If we want to state syllable-structure, we must explicitly introduce the 
element ‘syllable’ into our linguistic description, and state its relation to other elements 
of the linguistic hierarchy; it is precisely this which Chomsky and Halle fail to do” 
(Fudge 1999: 378). For her part, Selkirk (1999: 328) insists that English provides 
particularly strong evidence of the view that “the syllable is a linguistically significant 
unit that must have its place in phonological theory.” For her it is essentially “[…] a 
hierarchical unit, an internally structured tree […].” In support of this same stance, we 
find Durand (1990: 198–199) affirming that “The syllable needs to be recognized as a 
unit” and showing that the Polish stress rule, whereby word stress is assigned regularly 
to the penultimate syllable, can be formulated quite simply by accepting the syllable as 
a unit without any reference to syllabic composition. 
More recently, Goldsmith (2011: 164) reminds us that “The syllable is one of the 
oldest constructs in the study of language, and most studies of phonology have found a 
place for the syllable within them.” The philologists of the nineteenth century were fully 
aware of the fact that the outcome of the development of sounds is largely dependent 
on their phonetic contexts, which might mean their position in the syllable rather than 
their adjacency to other sounds in the speech chain. The pronunciation of many English 
words like resist /riˈzɪst/, absolve /əbˈzɒlv/ and exert /ɪɡˈzɜːt/, for example, complies with 
Verner’s Law, which, given certain circumstances, predicts voiced fricatives syllable-
initially before a stressed vowel instead of unvoiced ones. Similarly, the Second Germanic 
or High German Consonant Shift, which probably took place cAD 500 and distinguishes 
High German from the Germanic languages of the Channel coasts and North Sea area, 
led to an allophonic split among the unvoiced plosives whose outcome depended on their 
position in the syllable. Thus the unvoiced alveolar plosive /t/ was affricated initially 
([ts]), but became a fricative ([s]) in coda position, as can be seen by comparing German 
zehn, aus and Wasser with English ten, out and water.
* mott@ub.edu
Linguistica_2017_FINAL.indd   229 12.3.2018   13:08:28
230
All this, however, is not to say that the concept of the syllable is indispensable in 
modern phonological theory. For example, Harris (1994: 45) in his description of the 
phonology of English argues that “[...] for our present purposes, we can manage quite 
happily without a syllable node.” In fact, abstract theories, like government phonology, 
on the whole, seem to adopt a word-level only (van der Hulst/Ritter 1999: 42).
The aim of this paper is, firstly, to underline the shortcomings of some earlier, 
purely linear approaches to the syllable, which, view the unit, roughly speaking, 
either in terms of peaks of prominence derived from the relative sonority of sounds 
or from the point of view of permissible strings of phonological segments and, 
secondly, defend Wells syllabic analysis as laid out in the LPD and Wells (1990), 
while at the same time pointing out the unresolved issues of this approach regarding 
phonetic correlates of morpheme boundaries, over and against the more abstract, 
theoretical proposals.
But firstly, evidence for the existence of the syllable will be considered: namely, 
speaker intuition, on the one hand, and phonological processes that derive from the 
composition of the syllable, on the other.
2  EVIDENCE FOR THE EXISTENCE OF THE SYLLABLE
Speakers are aware of the existence of syllables and are able to count them, or at least 
count the number of peaks of prominence that are assumed to mark syllables. Ask any 
native or non-native English speaker how many syllables there are in the words master, 
penetrate and sympathetic, and the answer will be two, three and four, respectively.
Sometimes, however, there is uncertainty because of the tendency in English to 
compress unstressed syllables to make words fit as closely as possible the trochaic 
rhythmic pattern (stressed syllable + unstressed syllable). Thus, med(i)cine, batt(e)ry, 
dec(o)rative, radio ([i] > [j]), influence ([u] > [w]) may each be pronounced with one 
syllable less if compressed. Note also that even native speakers of English are often not 
aware that in words like table and listen, which contain an unstressed liquid or nasal, 
there are two syllables, and not just one.
A further problem in the identification of syllables is the precise location of the 
boundaries (see 3.1 below). How should the English words master, rebels, pastry and 
extra be divided? Is it ma.ster, mas.ter or mast.er; re.bels or reb.els; pa.stry, pas.try 
or past.ry; ex.tra or ext.ra (or should we split the <x> into [k] + [s]?)? Moreover, can 
morphological structure be overridden in syllabic division? Wells (see 3.2.1 [1] below), 
for one, would say no; thus for him the distribution of the vowels and consonants 
follows the pattern lend.ing and writ.er, not *len.ding and *wri.ter?
Language users are not accustomed to establishing syllable boundaries as they do 
not normally need to do so, at least as far as spoken language is concerned. If they 
do need to make a division in a long word in writing, then they will usually take 
into consideration morphemic structure (e.g. friend+ship, deliver+ing) or simply rely 
on works of reference to answer the problem for them. But syllable structure must 
be part of a speaker’s phonological knowledge because a native speaker of English 
knows that certain clusters are not well-formed. For instance, there are no English 
Linguistica_2017_FINAL.indd   230 12.3.2018   13:08:28
231
words that begin with [lt], though this cluster is possible when final (fault), a fact that 
is part of native speaker insight.1
A second argument that could be considered to testify to the existence of the 
syllable is the fact that it is the domain of many phonological processes. For example, 
in French maman [ma'mɑ͂] ‘mum,’ the final [n] nasalizes the preceding vowel, but 
the intervocalic [m] does not do the same as it is an onset consonant (ma.man), not 
a coda consonant. There are many processes that affect syllable edges: devoicing of 
final obstruents in some languages (see footnote 2); aspiration of initial voiceless 
stops in English; Homorganic Nasal Assimilation in syllable-final position in English 
(bump, kind, sink), but not necessarily across syllable boundaries (witness moonbeam, 
gumdrop, gangbang and Sten gun). Glottalling, tapping and rhotacization of English 
/t/ cannot occur initially, only finally, a weak position which allows suppression of 
supralaryngeal content and noise release.2
Note also how word play underlines the essential reality of the syllable for speakers: 
a big pan might be fumbled as a pig ban, thus producing a syllable-initial segment 
exchange, but is unlikely to be pronounced jocularly as a nig bap, with an exchange of 
consonants occupying different positions in the two syllables.
3  THEORIES OF THE SYLLABLE
In this section I shall refer to some traditional interpretations of the syllable that have 
proved to be inadequate in one way or another and then point out how Wells’ proposal 
remedies some of these shortcomings. The first approach that I examine is one based 
on the somewhat elusive notion of sonority; the second is a putative universal principle 
which states that, where possible, consonants are preferably assigned to onsets rather 
than rhymes.
1 There are analyses that would assign the final consonant of a word like fault and also that of a 
word ending in a single consonant, like fall, to the onset of a degenerate syllable, i. e., one without 
an audible nucleus. For a discussion of the advantages of this approach, see Harris (1994: 75). At 
the same time, note that the coda, while commonly occurring in government style onset-rhyme 
theories, is not, however, recognized as a constituent in government phonology (van der Hulst/
Ritter 1999: 24).
2 That the rhyme (or more precisely, the coda, if you accept its existence) is a “weak” or non-
robust position is attested by many phenomena apart from those mentioned. For example, in 
many varieties of Spanish, /s/ is aspirated (or even lost) in this position. Moreover, there tend 
to be fewer final consonants than onset ones: Spanish has no final plosives. Some languages, 
like Russian, Polish, Slovene, German and Catalan, neutralize the voice opposition and allow 
only voiceless [–SON] consonants in this position. Goldsmith (2011: 190) literally refers to the 
coda as “a position of neutralization,” and the onset-coda assymetry is further underlined by 
Silverman (212: 27–28), who states that neutralization “is almost exclusively found word-finally, 
and almost never word-initially.” Restrictions on consonant types and combinations or on vowel 
nuclei + consonants are also found. For example, English /ŋ/ only occurs after short vowels; long 
vowels are followed by clusters only if they contain an alveolar segment (joint, miles, seemed, 
but not *claimp, *boonk).
Linguistica_2017_FINAL.indd   231 12.3.2018   13:08:28
232
3.1  The Sonority Theory
The Sonority Theory (ST) argues that syllables can be recognized by a peak of 
prominence flanked by troughs, the sonority of a sound being “its relative loudness 
compared to other sounds” (Giegerich 1992: 132); thus the sonority first rises, then a 
peak is reached, and, after that, it falls according to what Selkirk (1984: 116) called 
a Sonority Sequencing Generalization (SSG) (see also Hogg/McCully 1987: 34). On 
a seven point scale, the relative sonority values of the sounds of English would be as 
follows, the most open vowel, /ɑː/, having the greatest degree of sonority, and a plosive 
consonant the smallest degree of sonority:
Sonority Level Manner of Articulation English Phonemes
7   Vowel   iː, ɪ, e, æ, ɑː, ɒ, ɔː, ʊ, uː, ʌ, ɜː, ə
6   Glide   j, w
5   Liquid   l, r
4   Nasal   m, n, ŋ
3   Fricative  f, v, θ, ð, s, z, ʃ, ʒ, h
2   Affricate  tʃ, dʒ
1   Plosive   p, b, t, d, k, ɡ
While this theory allows us to establish the number of syllables in a word by identifying 
the sonority peaks, it fails to determine the syllable boundaries. For example, it is clear 
that the word terminate has three syllables, as there are three peaks of prominence:
  1 7 4 7 4 7 1
/ t ɜː m ɪ n eɪ t /
However, there is no indication as to which of the following representations reflects 
the true syllabification of the word:
 
/ tɜː.mɪ.neɪt / / tɜːm.ɪn.eɪt / / tɜːm.ɪ.neɪt / / tɜː.mɪn.eɪt /
Another problem with the Sonority Theory, at least as far as English is concerned, is 
the fact that English allows three-consonant initial clusters beginning with [s] and final 
consonant clusters ending in [s], as in the words sticks and sprints (but see footnote 3). 
However, [s] is more sonorous than the adjacent plosives, and therefore contravenes the 
Sonority Sequencing Generalization, which requires the outer segments of syllables to 
be less sonorous than the inner ones.
Finally, the Sonority Theory cannot inform us about syllable constituency in 
different languages, i.e. the rules of co-occurrence. For example, both [f] and [s] are 
voiceless fricatives and therefore have the same sonority value but, whereas initial [sn] 
is possible in English, [fn] is not; other languages may allow both. Again, the Sonority 
Theory fails to tell us that *[fljɔɪlnd] is not a legal syllable in English, in which all three-
consonant initial clusters begin with [s] (spread, street, scream, etc.), so *[flj-] is ruled 
Linguistica_2017_FINAL.indd   232 12.3.2018   13:08:28
233
out (flew is /fluː/, not */fljuː/ in Present-day English). Likewise, the rhyme *[-ɔɪlnd] is 
illegal since three-consonant codas are supposed to be preceded by short vowels (cf. 
fifths [fɪfθs] with five [faɪv] or fives [faɪvz]).3 
Another restriction that a language may impose concerns sonority distance. A 
language may specify the condition that adjacent segments must stand at a minimal 
sonority distance from each other (Goldsmith 2011: 177; Steriade 1982: 94). For 
example, the definition of an optimal syllable according to the SSG predicts that in a 
two-member onset, the first segment cannot be more sonorous than the second, as in 
play /pleɪ/, or great /ɡreɪt/. However, [ml-] and [tl-] do not clash with the SSG and yet 
they are not possible onset sequences in English. These sequences appear to violate 
two different language-specific constraints of English. The sequence [ml-] violates 
a language-specific sonority constraint: within two-member onsets, there must be a 
minimum sonority distance of 2 points on the sonority scale, so that [fl-] and [ɡr-], for 
example, are acceptable. The sequences [tl-] or [fm-], to take another example, are not 
allowed because the two adjacent members of an onset cannot be homorganic, unless 
the first one is an [s], as in stay. Neither [tl] nor [fm] conform to this rule: in [tl-] both 
consonants are alveolar; in [fm-] both consonants are labial. The combinations [tr] 
and [dr] are admissible in English perhaps because [t] and [d] are alveolar, but [r] is 
postalveolar, and therefore not homorganic strictly speaking, and they are separated by 
at least two points on the sonority scale. On the other hand, these sequences tend to be 
treated monophonematically as affricates, anyway (see 3.2.1 [2]).
The sonority distance requirement for onsets does not seem to apply to codas. 
This is in accordance with the Dispersion Principle proposed by Clements (1990) and 
recited by Goldsmith (2011: 177) as: “all things being equal, a language will preferably 
maximize sonority difference in the onset, but minimize it in the coda.”
One final setback to the Sonority Theory that I have only seen mentioned in 
McCully (2009: 83) is the fact that we also perceive syllable structure when syllables 
are whispered. To this it might be added that sounds are produced and perceived, in 
any case, not so much in terms of a sonority scale, but in groups or classes whose 
membership is defined by the acoustic properties held in common, whether these be 
related to manner or place of articulation (McCully 2009: 85).
Sonority has been considered in this section on the basis of traditional data, in which 
obstruents are placed at the bottom of the scale, being assumed to possess negligible 
sonority, a fact which excludes them from occupying the syllabic nucleus, but there are 
languages for which syllabic obstruents have been posited, among them the Salishan 
languages, in which you get long strings of obstruents (see Wells 2016: 109 and Durand 
1990: 209). However, this raises basic questions about the nature of the syllable.
3 “Superheavy” onsets and codas are not accepted in all theories. For example, regarding onsets, 
Harris (1994: 63) says “[…] it is now widely assumed that the imposition of a two-position limit 
on onsets is universal […].” Moreover, Harris (1994: 62–63) gives initial and final [s] special 
status and cites proposals regarding initial [s], not only in English, but also in other languages, 
like Spanish and Italian. Either the segment is adjoined to some higher node in phonological 
structure or it is a coda consonant preceded by an unrealized nucleus.
Linguistica_2017_FINAL.indd   233 12.3.2018   13:08:28
234
3.2  The Maximal Onset Principle
The Maximal Onset Principle (MOP), referred to by Kennedy (2017: 205) as the 
“onset preference,” assigns as many consonants to the onset of a syllable as permitted 
by the phonological constraints of the language in question. Thus England must be 
syllabified as [ˈɪŋ.ɡlənd], since Modern English disallows [ŋɡ] in syllable codas; 
similarly actress is syllabified as [ˈæk.trɪs] because English has no onsets beginning 
with [ktr]; [ˈækt.rɪs] is not possible either because [tr] is an affricate and, as such, 
cannot be split (see 3.1 above).
The main problem with MOP is that it sometimes leaves English short stressed 
vowels stranded in open syllables and thus requires recourse to an additional rule of 
ambisyllabicity. If the words winner, honey and bother cannot be divided as [ˈwɪ.nə], 
[ˈhʌ.ni] and [ˈbɒ.ðə] in accordance with MOP, then ambisyllabicity states that their 
intervocalic consonants must belong to both syllables. But this looks like a piece of 
phonological patchwork unless we admit the idea of adjacent syllables being seamless, 
since these consonantal segments are not articulated as geminates. This is the approach 
in Local (1990: 355), where ambisyllabicity is regarded as structure-sharing and an 
integral part of parsing well-formed syllables, rather than a resyllabifying process. This 
theory postulates ambisyllabicity not just with simplex coda + onset sequences but also 
in cases like mister, where the cluster [st] would be treated as ambisyllabic, as would 
the [d] of hardy [ˈhɑːdi] despite the preceding heavy syllable (1990: 362).
Further problems arise with the phonological form of certain words like petrol: [tr] 
is syllable-initial in English, but not syllable-final (though see Wells’ division in 3.2.1 
[2] below). Given these circumstances, if we apply ambisyllabicity, should the word be 
divided as pet-trol? Note also the anomalous cases in RP (SSB, GB) where an [r] has to 
be spread between two syllables in words like severity [sə.ˈver.rɪ.ti]. The problem here 
is that in the aforementioned variety [r] is not pronounced in coda position. Compare 
vision [ˈvɪʒ.ʒn], in which [ʒ] would be assigned to an uncustomary initial position as 
well as a coda.
Strong arguments in favour of ambisyllabicity were put forward by Kahn (1976), 
whose theory of the syllable is autosegmental, and Durand (1990: 217) admits that 
“There is [...] evidence that syllables should be allowed to overlap.” However, Selkirk 
(1999: 343) rejects this theory for not allowing the syllable to be viewed as having 
an internal structure or be represented as part of a higher-order prosodic tree because 
autosegmental entities are not arranged in a hierarchy with respect to each other: “In 
an autosegmental analysis, it is impossible to view the syllable as a structural unit of a 
fully ramified prosodic structure.” Selkirk (1999: 345) does, however, lend support to 
MOP: “In general, when a medial consonant or consonant cluster may be analyzed as 
either a coda or an onset according to BSC (= basic syllable composition), it is the onset 
analysis which prevails.” In addition to this, Kennedy (2017: 210) makes the important 
observation that “[…] an internal CC sequence can only be assumed as tautosyllabic if 
the same sequence is observed word-initially.”
Before we leave MOP, it must be remembered that many of the objections to it stem 
from analyses of English data, but many languages fit the theory perfectly. Languages 
Linguistica_2017_FINAL.indd   234 12.3.2018   13:08:29
235
with a rigid CV structure, for example, present no problems of analysis. “MOP is 
applied through rule ordering: Coda Formation follows Onset Formation and applies 
only to unsyllabified consonants” (Hayes 2009: 254). Where none exist, there can be 
no problems of assignment.
3.2.1 Wells’ (2008/1990) Proposal for Syllable Structure
Wells 2008/1990 includes a proposal for syllabicity which is expounded more fully in 
Wells 1990: 76–86. In summary, it puts forward the following essential points:
1) Syllable boundaries coincide with word and morpheme boundaries. Thus there 
is a difference between grey#tape and great#ape, and between Tai.pei (T’ai.pei) 
and type#A. Similarly, kisses, lending and sunny are syllabified as kiss.es, lend.
ing and sunn.y, not *ki.sses, *len.ding and *su.nny. 
 For morphemes to be recognized as such, they have to be synchronically real 
to native speakers, although certain words act as if they were morphologically 
solid: bedr.oom, teasp.oon, mi.stake. In the case of the word mistake, the 
syllabification is mi.stake because native speakers no longer recognize the word 
as made up historically of the negative prefix mis- plus the verb take meaning 
‘consider.’ Moreover, the alveolar plosive is unaspirated (see [i] below).
2) Affricates are not split: catch.ing, Rog.er, petr.ol, detr.i.ment, paltr.y, caldr.on. 
Thus the affricates of catching and Roger, [tʃ] and [dʒ], respectively, cannot be 
divided as *[t.ʃ] and [d.ʒ].
3) The core of Wells’ theory is that, where the previous two conditions can apply 
without clashing with the phonological constraints of English (e.g. e.nig.ma 
cannot be *e.ni.gma as [ɡm] is not a syllable-final cluster in English), the 
stress (whether primary or secondary) attracts as many consonants as possible: 
plént.y, ùnd.er.stánd, vúlt.ure /ˈvʌltʃ.ə/. 
4) If a consonant is flanked by two unstressed syllables, it syllabifies to the left: 
sán.it.y, stím.ul.us.
The proposal in Wells is based on the behaviour of allophones at syllable boundaries. 
Thus, cases of aspiration, tapping, pre-fortis clipping, r-allophony, plosive epenthesis, 
t/d elision, and (de)voicing occur depending on whether the consonants involved are 
part of an onset or a coda. 
The following examples illustrate each of the linguistic phenomena listed above:
i) Aspiration: that mistake is syllabified as mi.stake. is confirmed by the fact that the 
[t] is not aspirated and must therefore be preceded by [s] in the same syllable.
ii) Tapping: debt.or, de.ter. As the [t] of the first word can be tapped or glottalized, 
it must be in the coda of the first syllable; in the second word the aspiration of 
the [t] confirms its status as syllable-initial.
iii) Pre-fortis clipping: leak.ing v. Lee#King. If the vowel in the first syllable of 
leaking is relatively short as compared to that of the name Lee King, then it 
must be because of pre-fortis clipping, which proves that the [k] in leaking is in 
the first syllable.
Linguistica_2017_FINAL.indd   235 12.3.2018   13:08:29
236
iv) r-allophony: your rise v. your eyes. The initial [r] of rise is somewhat rounded 
in many speakers, whereas the linking [r] of your eyes, if pronounced, does not 
have this property and is therefore syllable-final.
v) Plosive epenthesis: ten(t)s.ing v. ten sing. Possible plosive epenthesis in tensing 
proves that the [n] + [s] must both be coda consonants. No such epenthesis is 
possible in ten sing, where the [n] and [s] are heterosyllabic.
vi) [t]/[d] elision: [t] and [d] can only be elided in a consonant cluster which is 
syllable-final. Thus the final [t] of last can be elided in last wit, but the initial [t] 
of the word twit, as in crass twit, cannot be omitted; likewise, the [d] of old is 
often elided, for example in the expression old man, but not the [d] in coal dust.
vii) Voicing can also supply some information on syllabification. For example, in 
a.pply the [l] is voiceless, so the preceding [p], which is voiceless, must be in 
the same syllable; on the other hand, the [l] in app.li.cat.ion is relatively voiced 
and cannot therefore have been influenced by the preceding [p]. This shows 
that the [p] must belong to the preceding syllable.
3.2.2  Notes and Comments on Wells’ proposal
The argument that constituency within the syllable is based, at least in part, on 
distributional or phonotactic considerations is not entirely new (see Selkirk 1999: 329; 
Pike 1967: 386–387), but the detail provided by Wells (1990: 76–86) in his article on 
syllabification and allophony is unprecedented. 
Regarding each of the four basic conditions for syllabification outlined in 3.2.1 
(1–4), I would add the following notes and comments
1) Syllable boundaries coincide with word and morpheme boundaries.
 Recognition of the coincidence of morphemes and syllables is, probably 
rightly, at variance with earlier, more simplistic theories of the syllable, 
like MOP or the contention expressed by Fudge (1999: 376) that “the /v/ of 
drive is syllable-final, whereas it is more natural to take the /v/ of driving as 
syllable-initial.”
2)  Affricates are not split. 
 This makes sense in view of the fact that [tʃ] and [dʒ] are generally treated 
as monophonematic, a fact that in metrical phonology can be represented by 
showing that two melodic units are linked to one skeletal position (Harris 1994: 
40). But a problem arises with [tr] and [dr]: LPD divides petrol as /ˈpetr.əl/ 
since [tr] is a voiceless affricate and the stressed syllable requires a consonant 
after the short vowel. Since the affricate cannot be word-final, this calls for 
a distinction to be made between word internal positions and word edges, 
though perhaps this is not a bad thing, as these positions do often have different 
constituency. For example, Spanish has coda [b] internally in submarino, but 
[b] is rarely word-final (only in foreign words like club).
3) Conditions 1 and 2 v. sequential constraints on sounds.
 This usually works, but note the problems with morpheme boundaries 
mentioned below.
Linguistica_2017_FINAL.indd   236 12.3.2018   13:08:29
237
 It is interesting that, as English speakers pronounce the Spanish name Pepe /
ˈpepe/ as /ˈpepi/ or /ˈpepeɪ/, and not */ˈpeɪpeɪ/ or */ˈpeɪpi/, they must therefore 
be treating the intervocalic [p] as a coda consonant required by the preceding 
short stressed vowel, rather than an initial one. 
4) If a consonant is flanked by two unstressed syllables, it syllabifies to the left.
 This syllabic division would appear to be realistic in view of the fact that 
rhymes are left-headed (for detailed reasoning, see Harris 1994: 149).
Regarding the exemplification relating to allophonic distribution in syllables 
expounded in 3.2.1 (i–vii), I would add the following notes and comments:
i) Certain short vowels can occur finally in unstressed syllables, so the division 
of mi.stake shown by the LPD is legal, as would be that of a.spire and 
a.skance, which also contain unaspirated plosives. Both schwa and the kIt 
vowel, which we have in the first syllable of mistake, are very common in 
unstressed open syllables, but other vowels can appear too: Wells (1990: 83) 
quotes the case of no.stalg.ic, with the lot vowel in the first syllable, saying 
that the [s] may be captured by the stressed syllable as there is no morpheme 
boundary after it.
(ii) Regarding tapping, note that it usually takes place before a weak vowel ([i], [u] 
and [ə], and sometimes [ɪ] and [ʊ]), so the [t] of latex and Utah, for example, 
which contain full vowels in the last syllable, is not tapped. However, there is 
no problem across word boundaries, even if the following vowel is strong, as 
in put up, get up and quite obviously (Wells 2014: 180). 
 An interesting case of possible [t]-glottalling can be found in the variable 
pronunciation of the surname or place name Entwistle (Wells 2014: 36). Those 
who glottalize the [t] are treating it as syllable-final; those who don’t are 
treating it as syllable-initial. Thus the name is syllabified in different ways by 
different people: either as Ent.wistle or En.twistle.
 Wells (1990: 81) observes that Americans may tap the [t] of quality but not that 
of politics. This he attributes to the fact that the final syllable of quality has a 
weak vowel (/ɪ/ or /iː/, according to him), while the vowel of -ics must count as 
a full vowel sufficient to outrank the preceding weak vowel and capture the [t]. 
Another way of looking at this would be to recognize -ics as a heavy syllable 
(and some phonologists would say it bears secondary stress because of this), 
although this argument will not work with the adjective politic, where the final 
syllable with only one consonant in the coda is not heavy, and yet this word 
behaves in the same way as politics as regards syllabification and absence of 
tapping.
(iii) This seems to provide adequate phonetic evidence to show that that the coda 
does exist in the face of proposals to the contrary (see footnotes 1 and 3). 
 
Paragraphs iv–vii provide further evidence for the onset-coda dichotomy and 
underline further the importance of phonetic features in syllable constituency. So it 
Linguistica_2017_FINAL.indd   237 12.3.2018   13:08:29
238
would appear that, on the whole, Wells has created a reliable way of assigning consonants 
to syllables. Nevertheless, it needs to be pointed out that the analysis is hampered 
slightly by the many cases where morpheme boundaries seem to be overridden. Pre-
fortis clipping in a number of adjectives ending in -ful, like awf.ul, caref.ul, tearf.ul and 
joyf.ul, as opposed to others, like law.ful, rue.ful, woe.ful and sorr.ow.ful is a case in 
point. With reference to the former, Wells (1990: 81–82) says that we are not dealing 
with one of the “synchronic, psychologically real morphemes,” and he contrasts awful 
‘terrible’ with awe-ful ‘full of awe,’ which does not show pre-fortis clipping. It remains 
to be seen whether all the words with the suffix -ful that undergo pre-fortis clipping are 
more lexicalized, less transparent to native speakers, and less “psychologically real,” to 
use Wells’ terminology, than those that do not. 
Even harder to explain than the previous cases are the ones like plent.if.ul, its 
synonym bount.if.ul, and pow.erf.ul, in which Wells’ analysis overrides the morpheme 
boundary and applies the rule that assigns to the coda a consonant standing between 
two weak-vowelled syllables (3.2.1 [4]).
To take another example, Wells believes that -ford, unlike some other morphemes, 
like -ton and -son, does behave phonetically as a separate morpheme, as in Cray.ford 
/kreɪ.fəd/, but it must be remembered that short stressed vowels will require coda 
consonants, so Staff.ord, for example, will accordingly be divided as /stæf.əd/. As 
Wells (1990: 82) himself admits, “In general, this whole area of presence / absence of 
phonetic correlates of morpheme boundaries is still far from fully explored.”
3  CONCLUSION
This assessment of various, sometimes competing, approaches to an analysis of the 
syllable has hopefully brought to light some of the strengths and weaknesses of each 
one, in particular as far as English is concerned. While exclusive recourse to the 
acoustic properties of this phonological unit fail to cast light on the identification of 
boundaries, the interpretations that do address this question, like the Maximal Onset 
Principle and John Wells’ proposal, still run into many difficulties, not all of which 
have been resolved. As far as English is concerned, Wells’ approach appears to be the 
most practical, as is attested by the observations he has made on the distribution of 
allophones in syllable onsets and codas. This view has the advantage of steering away 
from the less tangible, more abstract arguments adduced by government phonology, for 
example. Nevertheless, there is still more work to be done on the relationship between 
morphology and the division of words into syllables. 
I note that in the LPD Wells analyses discredit as dis.cred.it. However, in the 
American recording of this word I hear no aspiration of the [k] or devoicing of the 
[r], which would assume the syllabic division to be di.scred.it. On the other hand, 
Wells gives discriminate as di.scrim.i.nate, but I believe that I personally devoice 
the [r], which would suggest that I assign the [s] to the coda of the first syllable. 
What are we to make of cases like these? I would go so far as to suggest that syllabic 
division in theory may clash with syllabic division in reality and maintain that, even 
though syllabification is predictable in most languages, there is a strong subjective, 
Linguistica_2017_FINAL.indd   238 12.3.2018   13:08:29
239
idiosyncratic element in the morphology of some languages, like English, that has 
hitherto been underrated, especially by largely abstract notions of linguistic analysis 
that seek to establish hard-and-fast rules for language behaviour. Here the role played 
by etymology in determining syllabification for individual speakers cannot be ignored. 
The more transparent a form is to the speaker, the more likely it is that he or she will 
divide it according to its historical components. Thus I have no qualms about assigning 
the [s] of dismember to the first syllable, in view of the prior existence of the noun 
member, but disparage is less clear to me. Even though dis- in this word is historically 
a separate morpheme, I feel inclined to analyse it as di.sparage. I end with some words 
relevant to this situation from Zec (2007: 191): “Constraints that make reference to 
morphological constituencies may affect the place of syllable boundaries […], while 
constraints on phonological constituencies higher or lower than the syllable may affect 
syllable shapes or their weight properties […]. It is left for future research to establish 
the range of such interfaces.”
References
CLEMENTS, George N. (1990) “The Role of the Sonority Cycle in Core Syllabification.” 
In: John C. Kingston/Mary E. Beckman (eds), Papers in Laboratory Phonology 1: 
Between the Grammar and the Physics of Speech. New York: CUP, 283–333.
DURAND, Jacques (1990) Generative and Non-linear Phonology. London: Longman.
FUDGE, Eric (1999 [1969]) “Syllables.” In: John A. Goldsmith (ed.), Phonological 
Theory. The Essential Readings. Oxford: Blackwell, 370–391.
GIEGERICH, Heinz J. (1992) English Phonology: An Introduction. Cambridge: CUP.
GOLDSMITH, John A. (22011) “The Syllable.” In: John A. Goldsmith/Jason Riggle/
Alan C. L. Yu (eds), The Handbook of Phonological Theory. Oxford: Wiley-
Blackwell, 164–196.
HARRIS, John (1994) English Sound Structure. Oxford: Blackwell.
HAYES, Bruce (2009) Introductory Phonology. Oxford: Wiley-Blackwell.
HOGG, Richard/Chris B. McCULLY (1987) Metrical Phonology: A Coursebook. 
Cambridge: CUP.
KAHN, Daniel (1976) Syllable-Based Generalizations in English Phonology. Doctoral 
dissertation. Cambridge, MA: MIT.
KENNEDY, Robert (2017) Phonology: A Coursebook. Cambridge: CUP.
KOHLER, Klaus J. (1966) “Is the Syllable a Phonological Universal?” Journal of 
Linguistics 2/2: 207–208. http://dx.doi.org/10.1017/S0022226700001493
LOCAL, John (1990) “Syllabification and Rhythm in Non-segmental Phonology.” In: 
Susan Ramsaran (ed.), Studies in the Pronunciation of English. London: Routledge, 
350–366.
McCULLY, Chris (2009) The Sound Structure of English: An Introduction. Cambridge: 
CUP.
PIKE, Kenneth (1967) Language in Relation to a Unified Theory of the Structure of 
Human Behavior. The Hague: Mouton.
Linguistica_2017_FINAL.indd   239 12.3.2018   13:08:29
240
SELKIRK, Elizabeth (1984) “On the Major Class Features and Syllable Theory.” 
In: Mark Aronoff/Richard T. Oehrle (eds), Language Sound Structure: Studies in 
Phonology Presented to Morris Halle by His Teacher and Students. Cambridge 
Mass: MIT Press.
SELKIRK, Elizabeth (1999 [1982]). “The Syllable.” In: John A. Goldsmith (ed.), 
Phonological Theory. The Essential Readings. Oxford: Blackwell, 328–350.
SILVERMAN, Daniel (2012) Neutralization. Cambridge: CUP.
STERIADE, Donca (1982) Greek Prosodies and the Nature of Syllabification. Doctoral 
dissertation. Cambridge, MA: MIT.
VAN DER HULST, Harry/Nancy A. RITTER (1999) The Syllable: Views and Facts. 
Berlin: Mouton de Gruyter.
WELLS, J. C. (1990) “Syllabification and Allophony.” In: Susan Ramsaran (ed.) 
Studies in the Pronunciation of English. London: Routledge, 76–86.
WELLS, J. C. (32008). Longman Pronunciation Dictionary. Burnt Mill (Harlow): 
Pearson/Longman. 
WELLS, John C. (2014) Sounds Interesting. Observations on English and General 
Phonetics. Cambridge: CUP. 
WELLS, John C. (2016) Sounds Fascinating. Further Observations on English 
Phonetics and Phonology. Cambridge: CUP. 
ZEC, Draga. 2007. “The Syllable.” In: Paul de Lacy (ed.) The Cambridge Handbook of 
Phonology. Cambridge: CUP.
Abstract
APPROACHES TO THE SYLLABLE: AN ASSESSMENT
In this paper, which looks back at some of the principal ways of viewing the syllable 
that have been proposed and attempts to assess their relative validity, I firstly refer to 
evidence for the existence of the syllable and lend support to the argument that the 
syllable must be part of speakers’ phonological knowledge because, not only can they 
count syllables, but they also know what sound sequences are permissible in them 
in their native language. Moreover, in this same vein I also recall the fact that the 
syllable is the domain of many phonological processes (in English, note aspiration 
and the glottalling, tapping and rhotacization of [t], for example). Finally, I defend the 
analysis of English syllabification proposed by Wells in the Longman Pronunciation 
Dictionary over and against other traditional linear models and the more abstract non-
linear models, while recognizing that Wells’ hypothesis still leaves unresolved some 
issues involving the presence/absence of phonetic correlates of morpheme boundaries. 
Keywords: the syllable, linear and non-linear phonology, syllabification and allophony, 
J. C. Wells
Linguistica_2017_FINAL.indd   240 12.3.2018   13:08:29
241
Povzetek
ANALIZA PRISTOPOV K ZLOGU
Članek obravnava nekatera glavna stališča do strukture zloga in jim skuša določiti 
relativno veljavnost. Najprej se posvetimo obstoju zloga in podpremo trditev, da zlog 
mora biti del govorčevega fonološkega vedenja, ker le tako zna prešteti zloge in se tudi 
zaveda, kateri glasovi si lahko sledijo znotraj enega zloga v njegovem maternem jeziku. 
Spomnimo tudi, da se znotraj zloga dogodijo številni fonološki procesi (v angleščini so 
to na primer asimilacija, glotalizirani, vibrantni in rotični [t]). Na koncu potrdimo vel-
javnost Wellsove analize delitve na zloge v angleščini, kot je predstavljena v Longman 
Pronunciation Dictionary, ki je drugačna od tradicionalnih linearnih modelov in tudi 
bolj abstraktnih nelinearnih modelov. Pri tem ugotavljamo, da Wellsova hipoteza pušča 
nerešena nekatera vprašanja, kot sta na primer prisotnost oziroma odsotnost fonetičnih 
povezav z morfemskimi ločnicami.
Ključne besede: zlog, linearna in nelinearna fonologija, delitev na zloge in alofonija, 
J. C. Wells
Linguistica_2017_FINAL.indd   241 12.3.2018   13:08:29