02sz,   PDF
Péter Szigetvári

Syncope in English*

1 Introduction English abounds in lexical consonant clusters; their nature and constraints on their occurrence are treated extensively in the phonological literature. There is, however, also a possibility for creating morpheme-internal consonant clusters postlexically, by syncope (also referred to as schwa-deletion, sonorant desyllabi cation or compression). Data on the conditions of syncope are contradictory, perhaps dialect-speci c.1 Not being a native speaker of the language, I have little chance of relying on my own intuitions about English pronunciation, therefore in this paper I will accept the data presented in the LPD (Wells 1990).2 Accordingly, the label \English“ below is to be taken to mean \LPDese”. This variant of English is often referred to as RP, although the term appears to be less and less justi ed (cf. the title of Varga’s (2002) interview). As for the tempo|an issue very relevant in any consideration of syncope | Wells explicitly excludes what he calls casual speech from the pronunciations he records (1990 : 241). I will begin with the issue of syllabic consonant formation (SCF). The inclusion of syllabic consonants in a discussion of syncope is not immediately obvious. In English, synchronic/dynamic syncope is always preceded | historically/derivationally|by SCF, for any C 1 @C2 C1 C2 alternation there exists an intermediate C1 C 2 stage; the reverse, as we are going to see, does not hold. This fact points to a strong connection between CC and syncopecreated consonant clusters.

• My thanks are due to Ádám Nádasdy and the editor for comments. As for mistakes

that remain: I could not always be convinced. 1

Cf. the rather different conclusions Kürti (1999) draws based on American data from Hooper (1978).

2

John Wells says “[compression] depends on which word we’re dealing with.” About the ungrammaticality of *“meldi (melody) he says “there is some constraint. Whether it’s to do with the word or something about l space d I’m not sure” (Varga 2002). It can be concluded that Wells is giving his transcriptions without any well-defined theory about the conditions of syncope, therefore they represent some kind of native speaker intuition, not data dictated by theory.

T H E E V E N Y E A R B O O K 5 (2002) 139–149

ISSN 1218–8808

139

140

Peter Szigetvari

Syncope proper is treated next, comparing it to SCF. High vowel gliding is also mentioned, since it seems to be a special case of syncope. The last but one section brie y treats apparent cases of syncope that do not t the patterns discussed previously. The aim of this paper is not to provide a theory to account for the data, rather to systematize the data to a certain extent in order to ease the task of further researchers, probably including myself.

2 Syllable reduction English is claimed to be a stress-timed language, one relevant consequence of which is that unstressed syllables reduce. This reduction process has several stages: rstly, the wide range of vocalic contrasts possible under stress shrinks to a meagre set of three, @ I U (this stage will be referred to as \3V”; cf. (1a));3 secondly, this set reduces to @, (1b)|in the case of nonhigh vowels, e.g., separate , the 3V and the @-only stage are homophonous. The loss in the prominence of the syllable is furthered by SCF, i.e., the loss of the vowel with the retention of its syllabicity, (1c). Finally, syllabicity may also disappear, which in traditional terms is equivalent to the total loss of the unstressed syllable, (1d). The four stages are illustrated by possible pronunciations of separate, family and natural. (1)

separate family natural

a. 3V set

sep@r@t fæmIli næÙUr@l

b. @-only

sep@r@t fæm@li næÙ@r@l

c. syll. C

sepr @t fæmli næÙr @l

d. syncope sepr@t fæmli næÙr@l

The four stages identi ed in the gradual reduction of unstressed syllables are not all simultaneously present: Wells does not give næÙUr@l as a variant of natural, for example, apparently he considers it obsolete in the dialect he records.4 Thus when I claim that some process like SCF or syncope is possible, this does not mean that it indeed occurs (and/or is recorded by

3

Wells transcribes some unstressed vowels by i or u. These do not increase the three-way distinction mentioned here, since i and u never contrast with I and U, respectively.

4

The variant does, however, occur in the EPD13 (Jones 1967), together with the other forms shown in (1).

Syncope in English

141

Wells) in all the items that match the given criteria. As most optional processes, syncope is also subject to variations due to idiosyncratic properties of lexical items, like frequency of use. Several interesting conditions govern the reduction of the 3V syllable to the @-only syllable (in fact, in some environments even the 3V type is impossible and the syllable retains a full vowel; more on these conditions in Nadasdy 1994). Here we will be concerned only with the loss of the schwa, \ rst“ resulting in a syllabic consonant, \later” possibly in syncope.

3 Syllabic consonant formation As Wells (1990 : xvii) notes, \a syllabic consonant always has an optional variant involving @ and a non-syllabic consonant.“ Consonants eligible for syllabicity in English are only the sonorants. The syllabicity (?) of fricatives, like in university ju:nIv3:s ti or dicult dIfklt, is (i) a fast speech phenomenon which I exclude from this discussion, (ii) not possible in LPDese and (iii) also peculiar in that, as opposed to most instances of syllabic sonorants, it occurs postvocalically. Among sonorants, although the glides j w are generally considered to be located between vowels and l in the sonority hierarchy, their syllabicity is radically di erent from that of l and nasals | r appears to pattern both with the latter group and with glides. To mention a few di erences: neither i, nor u alternates with @j or @w, in a way that l n m and @l @n @m do; both \syllabic” j and w (i.e., I and U) can occur stressed, while other syllabic sonorants cannot. (It must be admitted though that the high vowels that do alternate with their glide counterpart are never stressed anyway: j and w alternate with I and U, but not with
I and U
.) The nasals and l are also not equal in their inclination to turn syllabic. The liquid is best-known for this property, while of the nasals n is most frequent, m occurs less often and N“ is marginal.5 In fact, N” exclusively|given that the cluster @N is nonexistent in English |, and m potentially, evolves through place assimilation to the preceding obstruent: reckon rek@n ! rekn ! rekN“ ; 5

On the other hand, Bell (1978 : 169) notes, “syllabic nasals are greatly favored over liquids in the languages of our sample.” He lists 28 languages having both syllabic nasals and liquids, and 35 having only syllabic nasals, against a single language having only syllabic liquids. His list of languages may not be fully reliable, however, for example, Czech is claimed to possess both kinds of syllabic sonorants, but that language does not have syllabic nasals. Also, it does not take into account the lexical frequency of the two types of syllabic consonant in the languages. Still, the rarity of only-liquids languages is surprising.

142

Peter Szigetvari

open @Up@n ! @Upn ! @Upm. Nevertheless, this place assimilation is only possible if there is no vowel following: * rekNIN ” , *@UpmIN (cf. Wells 1990 : xxi). The availability of a @Cson string is not in itself enough for SCF to take place. Nontrivially, the input string must normally be preceded by a consonant: syllabic consonants do not occur word initially and only very rarely postvocalically. The rst case is conspired against by the fact that a word-initial unstressed vowel in a closed syllable fails to reduce to schwa, as angelic ænÃ
elIk , antagonistic æntæg@n
IstIk show; thus SCF, which, recall, presupposes the @-only stage, is impossible. There are not very many examples for the potential input vowel (a word-initial unstressed vowel in a closed syllable) anyway. The case of unless @nl
es is an exception, but *nl
es is still nonexistent. The same holds true for open syllables, where schwa does commony occur: allow *laU, annoy *nOI. The likelihood of these forms is lessened by the fact that the syllabic consonant in them is followed by a stressed vowel, though this in itself is by no means an obstacle to SCF: capitalistic -tl
IstIk , modernistic m6dn
IstIk are possible forms. My impression is that pretonic SCF is only possible if a morpheme boundary of some kind separates the syllabic sonorant and the following stressed vowel. As our examples show this can hardly be claimed to be a word boundary though, especially since the vowel(s) of such a sux would not be stressed. Furthermore, in this case a nonictus syllable containing a full vowel does not count as stressed (as opposed to their status with respect to syncope, see x4): SCF in Catalan kætl
æn, analogue ænl
6g, etc. appears to be normal. More interestingly, SCF also regularly 6 fails to occur after a stressed vowel. The following could serve as potential input: eon i:@n I@n , but *i:n; jewel Ãu:@l ÃU@l , but *Ãu:l. It is clear that the reason is not to retain the number of sonority peaks in the word, since the lexically bisyllabic forms can be drawn into a single syllable, as the transcriptions given second show. SCF is, however, possible after an unstressed vowel, but I have only found cases invovling -u@l (never U@l!) becoming -ul, like in casual, ritual, sexual, usual. Ignoring the forms discussed last, we may conclude then that for SCF the required input string is C@Cson . However, the nature of the rst consonant is also of importance in the process as the divergent grammaticality of camel kæml and column *k6lm demonstrates. The received wisdom in this respect is that SCF is possible in a C 1 @C2 string only if C1 is less

6

A few exceptions I have found: ideal aId
:l; Wells also has SCF in the non-RP and GenAm pronunciation of theorem T
:rm  , theorist T
:r@st.

Syncope in English

143

sonorous than C2 .7 The impossibility of SCF in both lemon and venom indicates that we must apply a broad sonority hierarchy, which treats all nasals alike, not one that distinguishes the sonority of sounds according to their place of articulation as well, like, e.g., Ladefoged’s (1982). If we accept the sonority condition, the behaviour of r turns out to be ambiguous (cf.  & Torkenczy 1986). While in C2 position it looks more sonorous than Acs either l or the nasals (celery selri, scenery si:nri, camera kæmr@), in C1 position it behaves as if less sonorous than any of the three ( barrel bærl, barren bærn, quorum kwO:rm). The duality of r is even more apparent in forms like arbitrary A:bItrr i, temporary temprri, or even temprri. Such behaviour is unparallelled by other syllabic consonants: consonant k6nsn@nt , but *k6nsnn t; similarly sIml@li , but *sImlli. In fact, it is not only r that behaves as an obstruent in C1 position, but also the glides j and w: loyal lOIl,8 equal i:kwl, narwhal nA:wl, GenAm year jr, lawyer l
6:jr.9,10 It is clearly not accidental that it is exactly these three approximants ( r j w) that exclusively occur before a syllabic segment|vowel or syllabic consonant|in English. 11 It is not only a di erence in sonority that is required between the two consonants anking the schwa in SCF, but also one of place of articulation. This claim is even less valid than the previous: it basically holds only of labials. None of the members of the short list of words containing a wordinternal Clab @m string12 exhibit SCF; word nally this string does not occur. 7

Under this condition postvocalic SCF should be impossible because the vowel (!) in the “C1 ” position is way too sonorous.

8

In light of the fact that SCF does not generally occur postvocalically, I take this form to be lOIjl; it must be admitted, on the other hand, that SCF does not occur after aI, though — perhaps somewhat inconsistently — Wells writes portrayal pO: “treI @l, i.e., pO:tr
eIl among other variants. The peculiarity of words in -ul may be explained along these lines: they actually end in -uwl, but then why don’t we have Ã
u:l for jewel ?

9

English does not abound in j@l, j@n or j@m sequences, which could serve as input for this case. In the handful of words with w@n, however, Wells does not show the possibility of SCF.

10

Without a clear idea of what the syllabicity of the high glides means, it is difficult to judge whether their sonority rank is also ambiguous with respect to SCF.

11

Although certain traditions of transcription seem to falsify this claim by giving forms such as meyk (make) beside yes (yes), lawd (loud ) beside wet (wet), or bArk (bark ) beside red (red ), distributional facts strongly argue that the nonprevocalic glides in such words are an integral part of the vowel, and are not consonantal “codas”.

12

blasphemous, -phemy, dismemberment, euphemism, infamous, infamy, sophomore

144

Peter Szigetvari

(Recall, word nally pm and its ilk are acceptable as a result of assimilation, like in open. ) Whether this is accidental, or implies some general property of SCF is unclear to me. The lack of this constraint for postalveolars ( natural -Ùr-, bene ciary -Sr-), the scarcity of words with labials in the relevant con guration and most importantly the fact that cross-linguistically \one favored source for syllabic consonants is C 1 VC2 where C1 and C2 are identical or at least homorganic” (Bell 1978 : 166) argue for the rst option. Stress also plays some role in the likelihood of SCF: the process is most frequent nonprevocalically, less likely before an unstressed vowel, and least likely, but still occasionally possible before a stressed vowel (cf., for example, Lebanon l
eb@n@n l
eb@n6n l
ebn @n, but *l
ebn 6n; megaron m
eg@r6n , but

• m
  egr6n vs. megara m
  eg@r@ m
  egr @). Without hoping to fully complete this

account of syllabic consonant formation, we now proceed to an examination of syncope.

4 Syncope We have seen that in English syncope presupposes a previous stage involv & ing syllabic consonant formation (this is the source of the problem Acs Torkenczy (1986) discuss). It follows then that the conditions of SCF are all necessary for syncope too, i.e., the latter is only possible if all the conditions of SCF apply.13 Syncope, however, is subject to further constraints: not all syllabic consonants can be desyllabicized. One crucial di erence between the two processes is that syncope only occurs if the sonorant to be desyllabicized is followed by an unstressed vowel.14 (Due to independent phonotactic constraints of English this vowel can only be @ or I/i, or, of course, a syllabic sonorant.) Accordingly, a word with an sww syllabic pattern may shorten to sw, while sws may not 13

Note that it is not universally necessary that a syncopated vowel be unstressed, cf. French ap
El ∼apl
e (appelle∼appeler ‘call-1/2/3sg.∼inf.’), Polish pj
Es∼ps
a (pies∼psa ‘dog-nom.∼gen.’).

14

A reviewer advises that “vowel” should be changed to “syllable”. I retain the first term (and mark stress accordingly), however, for several reasons. Firstly, any desyllabicized sonorant is followed by a vowel — otherwise syncope is not possible —, thus it becomes the onset of the unstressed syllable. Secondly, stress does not seem to be a property of syllables, but of vowels: the consonants of a syllable have practically no influence on whether that syllable can bear stress or not, but its vowel does, what is more, consonants, at least in English, never bear stress (perhaps r excepted in rhotic dialects). Thirdly, syllables are too theory-specific objects, vowels are not.

Syncope in English

145

shorten to ss (where s stands for a strong syllable, that is, one with a full vowel, w stands for a weak syllable, with a reduced vowel): memory m
em/ @ri vs. memorize *m
em/ @raIz . The generally accepted explanation for this regularity makes reference to notions like stress clash, or the preference of polysyllabic feet to monosyllabic ones: while both (memory) and (mem/ory) are polysyllabic feet, *(mem/o)(rze) would be two monosyllabic feet, at least the rst of which can be avoided by blocking syncope (Burzio 1994 : 61). This explanation is inadequate. If this were the reason for the absence of syncope before stressed vowels, we would expect this process to occur in swws strings, but this expectation is frustrated, as the following show: methodological *meT@d/@l
6ÃIk@l , hullabaloo *h2l@b/@l
u: . This, however, is not the result of some ununderstood ban on swws becoming sws , cf. nationalize næS@n@laIz
, næSn@laIz
, *næS@nlaIz
. In these words the only factor that inhibits syncope appears to be the stressedness of the following vowel. The correlation between SCF and syncope is most clearly demonstrated by the fact that the latter is only possible before sonorants (cf. Agatha

• ægT@

, sycophancy *s
Ikf@nsi ; counterexamples like veg/etable will be brushed aside in x6). It also holds that the consonant preceding the syncope site must be less sonorous than that following it (cf. family fæmli
vs. colony

• k
  6lni , enemy *
  enmi15 ). As we have seen, r and the high glides (j w) behave

as if less sonorous than l and the nasals (n m): the latter may turn syllabic when preceded by the former. At the same time, r can itself become syllabic when following l n or m. The same ambiguity does not hold in the case of syncope, as the data in (2) show. (2) a. tolerant t
6lr@nt general Ã
enr@l camera kæmr@

b. perilous *p
erl@s barony *bærni
caramel *kærm@l

The pronunciation patterns given in (2b) should occur, if r were ambiguous in syncope too. However, it is very dicult to prove that they are impossible, since, recall, the fact that the conditions elaborated here allow SCF or syncope in a given word does not guarantee that it in fact occurs. I have, nevertheless, not found any case of syncope preceded by r.16 15

Note the isolated lexicalized word enmity though, which is frequently pronounced
emn@ti .

16

As for caramel, its usual pronunciation ends in -mel, in which syncope is out, but alternative forms -m@l/ml are possible and potential inputs to syncope.

146

Peter Szigetvari

This asymmetry in the behaviour of r with respect to SCF and syncope is readily explainable by the well-known constraint on the occurrence of r j w h: these consonants (the glides) must be followed by a syllabic segment, a condition satis ed in an rC , but not in an rC cluster (cf. footnote 11). We have seen that as regards the identity of the preceding consonant SCF and syncope have similar requirements: in a C 1 @C2 string C1 must be less sonorous than C2 . However, while SCF is practically insensitive to the kind of consonant cluster that precedes the alternation site, (3a), syncope is not, (3b). (3) a. mantle mæntl
 chieftain Ù
:ftn angle æNgl
 patron sequel children mongrel rational Bracknel

p
eItrn s
:kwl Ù
Ildrn m
2Ngrl ræSnl
 ræSn
 l bræknl


b. gambolling oftener company

gæmblIN
6
ftn@ k
2mpni

patronage *p
eItrnIÃ equally

• 
  :kwli

sepulchrally *s@p
2lkrli centrally

• s
  entrli

rationally *ræSnli
signalling *s
IgnlIN

Syncope is only possible after a certain type of consonant cluster, all of which exhibit a falling sonority pro le and many of which occur word nally. 17,18 On the contrary, SCF may occur after just any cluster that occurs in the relevant position. This trivially shows that, on the one hand, syllabic consonants pattern with vowels. On the other, however, they share some characteristics with consonants, since their occurrence is sensitive to the sonority of the preceding consonant. If the claim made above that any syncope-created C 1 C2 cluster presupposes the existence of C1 C 2 , then the absence of word-initial syllabic consonants also entails that a word-initial schwa cannot be syncopated. Since the syncopation of an unstressed vowel is only possible before another unstressed 17

Wells has the GenAm datum patronal “peItrn@l “peItrnl. This is spurious, especially, since none of the other possible inputs (acronym, membranous, patronage, synchronous, synchrony) exhibit syncope.

18

The type of cluster after which syncope is possible in English is not easy to capture. One may be tempted to say this is the type that occurs in branching codas, however, some (e.g., mb, sÙ, etc.) do not occur word finally. What comes closest to the class is government phonology’s coda–onset clusters.

Syncope in English

147

vowel, the fact that words cannot begin in English with two unstressed syllables conspires against the syncopation not only of word-initial schwas, but of any schwa in the rst syllable of a word.

5 High glides Wells (1990 : 152f) uses the term compression for the process I have referred to as syncope here. In addition, he applies it to an apparently di erent phenomenon, to the drawing together of two syllables in words like medial m
:di@l m
:dj@l,19 where the rst pronunciation is three syllables long, the second only two, let us call this, not unexpectedly, (high vowel) gliding. One can attempt to interpret this loss of a syllable peak in the same way as sonorant desyllabicization. What has to be hypothesized is the nonexistent form *m
:d@j@l from which syncope regularly produces the attested m
:dj@l . In (4) I o er forms that look parallel. (4) medial family

”

mi:di(j)@l m:idj@l