Vowel Clipping in English

PTFonR05c.qxp
2005-08-24
22:29
Page 77
Vowel Clipping in English
Skracanie samog³osek angielskich
Wiktor Gonet1, Lidia Stadnicka2
Department of Phonetics and Phonology, Institute of English,
Maria Curie-Skłodowska University
[email protected]
2
OSPHON Seminar*, Maria Curie-Skłodowska University
[email protected]
1
ABSTRACT
It is well known that English fortis consonants shorten the duration, or ‘clip’, the
preceding vowels. Most sources describe this phenomenon en block, not making
distinctions concerning speech tempo or vowel tenseness. The present study makes
an enquiry into the conditioning of clipping with regard to intrinsic vowel duration
in two speaking modes: monosyllables and connected speech. It is found out that
clipping operates in a statistically significant manner in monosyllables, especially in
tense vowels and /Q/, and less so in the remaining lax vowels. In connected speech
only two tense vowels and /Q/ are clipped with statistically significant force, while
in the remaining vowels the variability range due to clipping is statistically nonsignificant.
STRESZCZENIE
Powszechnie wiadomo, że angielskie spółgłoski mocne skracają iloczas poprzedzających je samogłosek. Większość źródeł opisuje to zjawisko w sposób blokowy, nie
wnikając w zmienność związaną z tempem mowy lub iloczasem podstawowym samogłosek (‘opozycją długa-krótka”). Obecna praca przedstawia uwarunkowanie
skracania od iloczasu podstawowego samogłoski w dwóch typach wypowiedzi:
w monosylabach i mowie wiązanej. Wykazuje się, iż skracanie operuje w sposób
istotny statystycznie w monosylabach, szczególnie w odniesieniu do samogłosek
długich i /Q/. W mowie wiązanej tylko dwie samogłoski długie i /Q/ skracają się
w sposób istotny statystycznie; w pozostałych zmienność nie jest istotna.
1. Intrinsic and extrinsic vowel duration
The concept of ‘vowel duration’ of English vowels is ambiguous because it can refer
either to intrinsic or to extrinsic durational variability.
Intrinsic variability concerns the popular ‘short/long’ distinction, in the IPA
symbolised by the use of the colon (˘) for the ‘long’ vowels, and its lack for the short
vowels. This variability, although historically phonemic, has lost its contrastive power
*Open Seminar in Acoustic Phonetics run by dr. Wiktor Gonet
77
PTFonR05c.qxp
78
2005-08-24
22:29
Page 78
Speech and Language Technology. Volume 8
and functions as a factor enhancing the more basic qualitative contrasts. Thus the use
of this graphic convention can be misleading for the foreign adept of English
pronunciation as it deemphasizes the importance of the qualitative contrasts; moreover, its use, restricted to monphthongs, separates them from diphthongs.
There is another, weaker intrinsic conditioning of vowel duration related to their
degree of openness: the more open the vowel, the longer it is. There have been attempts
to explain this relationship on jaw inertness grounds by claiming that the jaw needs
more time to travel to the open position than it does for the close position (Lehiste,
1970; Cochrane, 1970). Yet it is argued elsewhere (Lindblom, Lubker and Gay, 1979;
Condax and Krones, 1976) that this effect can be neutralised by compensatory motor
mechanisms; Gonet (1989) formulates a hypothesis that the motivation of this
lengthening lies in the desire to enhance the perception of vowel quality.
Extrinsic variability in vowel duration refers first of all to the well known
shortening, or ‘clipping’ (Wells, 1995, p. 136) effect exerted on vowels and diphthongs by postvocalic fortis obstruents. This effect also extends onto sonorants if they
occur between the vowel and the obstruent; thus, for instance, both the vowel and the
lateral are shortened in [melt]. Vowels followed by lenis obstruents and sonorants
preserve their duration, and so do the ones in open syllables. It has been convincingly
argued that extrinsic duration variability has the function of indicating the voicing of
the following consonant (Walsh and Parker, 1981).
The duration of a vowel followed by an obstruent is further modified by the
operation of another force, i.e. manner of articulation of the following obstruent. Thus
vowels are shorter before stop consonants (plosives and affricates) and longer, before
fricatives (Gonet 1989, 1997).
Finally, the duration of vowels is further modified by suprasegmental conditioning;
it is inversely related to the number of syllables and directly proportional to the degree
of its stress (Jassem 1971).
There is general reluctance to use the ‘short/long’ designation of the intrinsic
variability because of the relativity caused by the fact that extrinsic conditioning can
override intrinsic duration so that the unreduced ‘short’ vowels are longer than the
shortened ‘long’ vowels (Gonet 1989, 1997). Two other cover terms are used in
literature: ‘checked/unchecked’, advocated by Jassem (1971) and Vasilyev (1970), and
‘lax/tense’, reflected directly in the SPE (Chomsky and Halle, 1968) distinctive feature
framework. Both have the advantage of treating monophthongs and diphthongs as a
natural class. The feature-based distinction is used more often, despite serious
ontological problems (Lass, 1976). Hence, the designation lax/tense will be used to
refer to the intrinsic durational variations, while ‘clipped/full length’, to the extrinsic
variability induced by the strength of articulation of the postvocalic consonant.
PTFonR05c.qxp
2005-08-24
22:29
Page 79
Vowel Clipping in English
79
2. Vowel duration in isolated words
Gonet (1989, 1997) studied the conditioning of vowel duration by three intrinsic (tensenss, openness, advancement) and three extrinsic (force, manner and place of articulation of postvocalic consonants) factors; the study was based on 648 measurements of
monosyllabic words spoken by 4 speakers (2 male and 2 female). It was shown that by
far the strongest factors conditioning vowel duration are tenseness and pre-fortis
clipping, the range of durational variability of vowels in monophthongs (tense vs. lax)
being the same as the scope of variability conditioned extrinsically by the force of
articulation of the post-vocalic consonant and equals 1,5:1,0; consider Fig. 1:
Fig. 1. Duration of vowels in monosyllables: main effects.
In most previous studies, force of articulation was also found to be the strongest
extrinsic effect determining vowel duration (House and Fairbanks, 1952:128; Peterson
and Lehiste,1960:200; and House 1961:374), while in one source (Imiołczyk 1974:41)
it ranks second to tenseness. In Gonet (1989, 1997) the strength of both factors is
equal, as the proportions of duration in both main effects are almost the same. In terms
of statistical inference, force of articulation is significant at α=0.001, while tenseness,
at α=0,01. The duration of the increments are in both cases around 35% , while the
slightly rounded proportions between the levels are 100 : 65.
A more interesting picture emerges when their effect is viewed as an interaction;
this is shown in Fig. 2:
PTFonR05c.qxp
80
2005-08-24
22:29
Page 80
Speech and Language Technology. Volume 8
Fig. 2. Duration of vowels in monosyllables: interaction.
In the tenseness x force of articulation interaction template, there are three statistically
different levels (α=0.001):
Tense Lenis > Tense Fortis = Lax Lenis > Lax Fortis
and clipping occurs both in tense (100:60) and lax (100:70) vowels, with clipped tense
vowels having the same duration as full length lax vowels.
PTFonR05c.qxp
2005-08-24
22:29
Page 81
Vowel Clipping in English
81
3. Vowel duration in connected speech
The development of speech synthesis and the desire to improve its quality prompted
the need of checking how far the conditioning of vowel duration can be applied to
running speech. With this view in mind, another vowel duration project was developed
(Bąk, 2000) in which vowel duration was studied in the speech of five native speakers
of English. Thje conclusions confirm Gonet’s findings as to the interplay of intrinsic
and extrinsic durational variability (Gonet, 1989, 1997), although the ranges of
variability are smaller 2,0:1,0 for the tense/lax contrast, and 1,0:0,8 for the extrinsic
conditioning. Similarly, the significance assessments are weaker.
Let us first consider the main effect data for connected speech in Fig. 3:
Fig. 3. Duration of vowels in connected speech: main effects.
In both cases the differences between the means are significant at α=0,001. It can
be noticed that the difference between the means is much more articulate for tenseness
(45%) than it is for clipping (15%), while the rounded proportions of the levels are
100:55 for tenseness and 100:85 for clipping. Let us now view the interactive effect
of tenseness and clipping, significant at α=0,001 shown in Fig. 4:
PTFonR05c.qxp
82
2005-08-24
22:29
Page 82
Speech and Language Technology. Volume 8
Fig. 4. Duration of vowels in connected speech: interaction.
In the tenseness x force of articulation interaction template for connected speech,
there are three statistically different levels (α=0.001):
Tense Lenis > Tense Fortis > Lax Lenis = Lax Fortis
where the rounded proportions of duration are 100:75 for tense, and 100:90 for lax
vowels. The picture emerging from these data presents a situation different from the
interaction obtained for monophthongs (cf. Fig. 2). Despite the fact that here, too,
three durational classes emerge, the extrinsic variability in lax vowels is neutralized;
apparently because of the relatively short duration of the lax vowels themselves; in
addition, the clipped tense vowels are longer than the full length lax vowels (in
monophthongs they were of equal duration).
4. Clipping in monosyllables and connected speech
Combined results of clipping for two speech modes are shown in Fig. 5:
Fig. 5. Combined means for monosyllables and connected speech
(T*=tense, L*=lax, *F=fortis, *L=lenis, asterix ‘*’ is a variable).
PTFonR05c.qxp
2005-08-24
22:29
Page 83
83
Vowel Clipping in English
It becomes obvious that, when one passes from more considerate, slow speech to less
exact informal enunciation, the extrinsic durational relations are copied but reduced in
the force of the effect. In short, in monosyllables, clipping takes place in both tense and
lax vowels, while in connected speech, only tense vowels are clipped.
5. Clipping in individual vowels
The data on which the above conclusion was formulated were averaged across all
monophthongs; it would be interesting to see how far individual vowels undergo
clipping in both modes of enunciation.
5.1. Monosyllables
In monosyllables, the differences between the durations of clipped and non-clipped
vowels are significant at α=0,001 for 10 out of 11 vowels; there is no statistically
significant clipping effect for /U/; this is shown in Figure 6. It can be seen that the
extent of clipping varies from extensive for high tense vowels, less extensive for the
remaining tense vowels and /Q/, to relatively small but significant for the lax vowels
except /U/ in which the effect is negligible.
Fig. 6. Clipping in monosyllables.
Table 1. Key to phonetic symbols
1
2
3
4
5
6
7
8
9
10
11
A˘
Œ˘
ç˘
u˘
i˘
Q
Å
√
e
I
U
PTFonR05c.qxp
84
2005-08-24
22:29
Page 84
Speech and Language Technology. Volume 8
2.2. Connected speech
The effect of clipping in connected speech is much weaker than in monosyllables; this
is shown in Fig. 7:
Fig. 7. Clipping in connected speech.
The plot in Fig. 7 shows that, while the differences obtained due to clipping in lax
vowels and most tense vowels are negligible, there are only three vowels that preserve
the significance at α=0,01 of the difference between clipped and non-clipped variants,
viz. the tense /i˘, ç˘/ and the lax /Q/. These three vowels hardly constitute a natural
class; one can venture a generalisation that in connected speech clipping is significant
for longer front vowels and the strongly rounded /ç˘/ that requires an extensive lip
rounding gesture. Additionally, for three vowels (/A˘/, /Å/ and /I/) the effect is reversed.
6. Clipping in foreign learners of English
Let us now correlate these findings with data concerning vowel duration and clipping
in texts produced by foreign learners of English and the perception of it by native
speakers of English.
The findings in Gonet and Pietroń (2004, in press) suggest that to secure
favourable native speaker impression it is particularly important for Polish learners of
English to preserve the correct intrinsic durational relations in the English front
vowels /i˘/-/I/ and /e/-/Q/. This correlates well with the findings pertaining both to
clipping in both monosyllables and connected speech, where the two vowels /i˘/ and
/Q/ belong to those in which the clipping effect is strongest.
PTFonR05c.qxp
2005-08-24
22:29
Page 85
Vowel Clipping in English
85
Similar conclusions are reached in Szpyra-Kozłowska (in press), who presents the
results of a questionnaire filled in by 30 native teachers of English, who were asked
to evaluate the intelligibility of Polish speech and Polish accent. According to the
evaluators, making no distinction between long and short vowels, particularly between
/i˘/-/I/ and /e/-/Q/ was considered a grave error; other pairs of lax/tense vowels, i.e. /Å/
and /ç˘/ /√/ and /A˘/, /U/ and /u˘/ were viewed as less problematic. Szpyra-Kozłowska
observes that “these judgements correspond closely to the number of existing English
homophones that utilize these contrasts. Clearly, the preservation of the /i˘/-/I/
distinction is essential for numerous minimal pairs (e.g. ship – sheep, sit – seat, pitch
– peach), with the /e–Q/ contrast also fairly crucial (e.g. bed – bad, set – sat, ten –
tan), while the neutralization of other differences, i.e. between /Å/ and /ç˘/ /√/ and /A˘/;
/U/ and /u˘/ rarely leads to the rise of homophones and either miscomprehension or
confusion. This, in turn, suggests the amount of attention that should be given to
practicing the specific vowels and vowel pairs.”
Both in Gonet and Pietroń (2004, in press) and Szpyra-Kozłowska (in press), the
incorrect devoicing of word-final obstruents was also judged to be a serious error
hindering the communicative and aesthetic correctness of speech. In the work of Gonet
and Pietroń is was shown that the accompanying lack of correct clipping made foreign
English speech less attractive in both aspects of evaluation.
Thus the speech of foreign learners of English and its evaluation by native
speakers provides an additional independent area to test the validity of the present
findings.
REFERENCES
Bąk, L.M. 2000. Duration of English Monophthongs in Connected Speech. Unpublished M.A.
Thesis. Lublin: Maria Curie Skłodowska University.
Chomsky, N., and Halle, M. 1968. The Sound Pattern of English. New York: Harper and Row.
Cochrane, G.R. 1970. Some vowel durations in Australian English. Phonetica 22:224–250.
Condax, J.D., and Krones R.R. 1976. Duration of four vowels in manually produced synthetic
speech. Journal of Phonetics 4, 151–171.
Gonet, W. 1997. Duration of R.P. Monophthongs in Monosyllabic Words. Speech and Language
Technology Vol. 1 pp. 195–228. Poznań: Polish Phonetic Association / Wrocław: Wydawnictwo
Prac Naukowych FORMAT. Edited by W. Jassem and Cz. Basztura.
Gonet, W. 1989. Factorial Analysis of the Duration of R.P. Monophthongs in Monosyllabic
Words. Unpublished Ph. D. dissertation. Lublin: UMCS.
Gonet, W., and Pietroń G. 2004. The Polish Tongue in the English Ear. Zeszyty Naukowe Państwowej Wyższej Szkoły w Koninie. Nr 1/2004(4), str. 56–65.
Gonet, W., and Pietroń G. (in press). English native speakers’ perception of Polish accent. Speech
and Language Technology 8. Poznań.
House, A.S., and Fairbanks G. 1952. The influence of consonant environment upon the secondary
acoustical characteristics of vowels. Journal of the Acoustical Society of America 25:105-114.
Reprinted in Lehiste, I (ed., 1967:128–136).
House, A.S. 1961. On vowel duration in English. Journal of the Acoustical Society of America
33:1174–1178.
PTFonR05c.qxp
86
2005-08-24
22:29
Page 86
Speech and Language Technology. Volume 8
Imiołczyk, J. 1979. Segmentation and duration of R.P. monophthongs in selected monosyllabic
words. Unpublished M.A. thesis. Wrocław: Wrocław University.
Jassem, W. 1971. Podręcznik wymowy angielskiej. Warszawa: PWN
Lass, R. 1976. English phonology and phonological theory: synchronic and diachronic studies.
Cambridge: Cambridge University Press.
Lehiste, I. 1967. Readings in Acoustic Phonetics. Cambridge, Massachussetts and London,
England: The M.I.T. Press.
Lehiste, I. 1970. Suprasegmentals. Cambridge: M.I.T. Press.
Lindblom, B., Lubker J., and Gay T. 1979. Formant frequencies of some fixed-mandible vowels
and a model of speech-motor programming by predictive simulation. Journal of Phonetics,
7:147–161.
Peterson, G.E., and Lehiste I. 1960. Duration of syllable nuclei in English. Journal of the Acoustical
Society of America 32, 693–703. Reprinted in Lehiste, I. (ed., 1967:191–201).
Szpyra-Kozłowska, J. (in press). Intelligibility versus Polish accent in English. Studia Phonetica
Posnaniensia.
Vasilyev, V.A. 1970. English Phonetics: A Theoretical Course. Moscow: Higher School Publishing
House.
Walsh, T., and Parker F. 1981. Vowel length and voicing in a following consonant. Journal of
Phonetics 9:305–308.
Wells, J. 1995. Pronunciation Dictionary (Eighth impression). Longman.