|
|
web-sls 96.01 20.09.96
|
|
Collins and Mees claim in their chapter on elision in The Sounds of English and Dutch (1984) that word final /k/ cannot be elided in English if it is preceded by a nasal and followed by a consonant in the next word. This experiment, however, shows that /k/ sometimes does get elided in connected speech. In some cases there is no detectable silence, which would correspond to the /k/, between the nasal and the following initial consonant.
Note
The brackets < > are used in this paper to demarcate symbols of a greater level of generality than phonemic; thus <n> represents all nasals, and <C> represents all consonants.
In the chapter on elision in the textbook The Sounds of English and Dutch by Collins and Mees (1984) the authors point out to Dutch learners of English that they should not elide final /k/ following <n> if it is followed by another consonant, since 'this is not permissible in English' (Collins and Mees, 1984). This is quite a strong claim to be made about an elision in one particular environment. It could well be the case that /k/ is hardly ever elided in this environment, but there might be instances where this elision does not result in an unacceptable pronunciation. The aim of this experiment then, is to look at some natural speech data and to see whether final /k/ is always retained if it is preceded by <n>, and followed by another consonant.
The assumption that final /k/ is never elided if it is preceded by an <n>, and is followed by <C>, is neither supported nor refuted in the literature. Both Jones (1962) and Gimson (1980) mention elision in English. Both authors describe what elision is, and give the most common examples. Neither author mentions the elision of /k/ in this particular environment. They do refer to the more common elision of final /t/ in the same environment. The fact that neither author mentions the elision of /k/ suggests that it is not a common form of elision. This does not entail, however, that they support the claim made by Collins and Mees (1984). In order to test this claim a small experiment was carried out to examine possible instances of /k/ elision and test the Collins and Mees claim.
In order to test Collins and Mees' (1984) claim that final /k/ does not get elided if preceded by <n>, and followed by a consonant, sufficient data has to be accessed and examined. This experiment makes use of the MARSEC database, which is a digitised and elaborated adaptation of the SEC (Spoken English Corpus) database (Roach et al. 1993). The SEC was collected and transcribed in the 1980's, and consists of nearly six hours of spoken RP English, mainly taken from radio broadcasts. To make the audio information of the database more easily accessible, the recordings have been digitised and stored on a CD-ROM disk. Because of the new format, more extensive analyses can be made, and pieces of recording are easier to find. With the aid of the XWAVES speech signal analysis package, it was possible to look at the particular environment of <nk# C> and make a spectrogram for each phrase. An example of the spectrograms used is shown in Fig.1.
Fig.1 Spectrogram of the phrase think you'd
|
|
With a special searching program it was possible to compile a list of occurrences of the format <nk# C> in the transcribed text. Thus, 67 instances of this particular environment were found in the database. From these 67 instances, however, a considerable number had to be ignored. Two phrases were omitted because both these phrases was stored twice under different file names in the corpus. So in both cases the exact same phrase appeared twice in the data. A further 31 phrases had to be taken out of the experiment because in these phrases the initial consonant following a final /k/ was a plosive or /ð/ (which is usually pronounced as a lax dental plosive rather than as a fricative). In connected speech, two adjacent plosives very often just show one complete closure on a spectrogram. Sometimes the first plosive is simply replaced by a glottal stop, but quite often both plosives are audible in speech. It is impossible however, to determine where one plosive ends, and the other begins. So in this experiment the phrases in which /k/ was followed by another plosive were ignored. The number of suitable phrases was thus reduced to 34.
All 34 phrases were located in the MARSEC corpus. These were then examined and a spectrogram was produced for all <nk# C> environments. The main characteristic of plosives is a short closure in the vocal tract. Due to this closure plosives are quite distinct on a spectrogram. During the closure, a spectrogram shows a clear blank area where no frequencies are detected. Directly after the closure a sharp onset and sometimes some noise from aspiration indicates the moment of release. In this experiment the presence or absence of a completely blank part followed by a sharp onset on the spectrogram was the decisive factor in determining whether a /k/ was pronounced or not. Other characteristics of /k/ that might survive elision, for instance velarisation of the following sound, where ignored in this experiment. Partly due to the fact that this was a small scaled experiment, but also because a closure in the vocal tract is the most important definition of a plosive. It was therefor assumed that the absence of a closure in the vocal tract meant the elision of /k/. Each spectrogram of the selected phrases was looked at individually, and in all the cases the duration of the closure in the vocal tract was measured.
It was important also to establish that in the <nt# C> environment, where C is not a plosive, the /t/ often does get elided. So nine further phrases were selected which all provided a suitable environment for the elision of /t/. The same procedure was followed as for the <nk# C> environment: the duration of the silent interval was measured on-screen with the aid of spectrograms.
After the selection of all the phrases that contained the <nk# C> environment had been made, the duration of the silent interval (which is in effect the closure phase of /k/) before the onset of C could be measured. All measurements were done on spectrograms generated by XWAVES. The durations of silent interval are represented in the barchart (Fig.2).
Fig.2 Results Barchart
|
|
After the silent intervals had been measured for all 34 instances of <nk# C> the average duration of the silent intervals could be calculated. The average time of silence was 33.47 ms with a standard deviation of 22.73 ms.
From the nine selected phrases containing the <nt# C> sequence, seven showed an elision of /t/. In these seven cases, no silence could be measured between the nasal and the following non-plosive, consonant. The other two phrases showed clear silences of 26 ms and 22 ms.
As the barchart (Fig.2) shows, there are four instances where no silent interval can be observed between the nasal and the initial consonant of the following word; the contexts were as follows: think much, think much of, thank you very and think you'd better. This indicates that there is no complete closure in the vocal tract, which would create a short silence. Since /k/ requires a complete closure, it must have been elided in the four instances mentioned above. This is in contrast with the claim of Collins and Mees (1984) that /k/ is not elided in this environment.
The results for the <nt# C> environment however, do support claims made by Collins and Mees (1984), Gimson (1980) and Jones (1962), namely that /t/ does often get elided in this context. In seven out of nine instances there is no measurable silence between the nasal and the initial consonant of the following word, which indicates an elision of the /t/. So here all authors agree. The seven instances where /t/ gets elided in a <nt# C> context do not show a uniform pattern. In two cases the final /t/ is elided when followed by an /f/ or by an /m/ and once each when followed by /s/, /w/ or /j/. So /t/ is elided before fricatives, nasals and approximants. In the case of /k/ in the same environment however, elisions only took place twice before an /m/, and twice before a /j/. Thus /k/ was only elided if followed by an approximant or a bilabial nasal. The data does not provide enough evidence to make any claims about restrictions on the context in which /k/ can be elided if preceded by a nasal and followed by a consonant. Nevertheless, it seems that /k/ is more likely to be elided if it is followed by an approximant or by a bilabial nasal.
The results of this experiment contradict the claim that /k/ is not elided in the context <nk# C>. Out of 34 phrases containing this context, four phrases show an elision of /k/; so in 11.8% of the cases, /k/ is elided. Therefore it seems inappropriate to say that 'this is not permissible in English'.
The question that remains unanswered in this experiment is why there are many cases where /k/ is not elided in the context <nk# C>, whereas /t/ in this context is elided rather frequently. Almost certainly it has to do with the different phonological characteristics of /t/ and /k/, but a further investigation is needed to give an answer to this question.
I wish to thank Gerry Latawiec, Simon Arnfield and Peter Roach for their help throughout the experiment, and for their help in writing this article.
Collins, B.S. and Mees, I. (1984) The Sounds of English and Dutch. Leiden: E.J. Brill / Leiden University Press
Gimson, A.C. (1980) An Introduction to the Pronunciation of English. (3rd ed.) London: Arnold
Jones, D. (1962) An Outline of English Phonetics. Cambridge: Heffer
Roach, P., Knowles, G., Varadi, T. and Arnfield, S. (1993) MARSEC: A Machine-Readable Spoken English Corpus. Journal of the International Phonetic Association 23:2, pp.47-54
After secondary school Erik went to Leiden University to study English. Right from the beginning, linguistics was his favourite topic. In his third year he went to Reading University for one year. In Reading he studied in the Department of Linguistic Science. For the course entitled Experimental Phonetics students had to do a small project; Erik choose to investigate the elision of /k/ in <nk# C> context.
Erik has finished the year at Reading, and is currently in his fourth year at Leiden. The Linguistics Department at Reading has made him so enthusiastic that he is now doing two courses; one in English, and one in Linguistics! He hopes to finish both degrees in 1998.
download black and white versions of the figures for monochrome printing
web-sls article: #96.01
published: 20th September 1996
archived: n/a
academic comments to: the author
publishing comments to: Mark Tatham, Technical and Publishing Editor
Your attention is drawn to the web-sls copyright notice on the Journal's front page.
