the voicedness of intervocalic word-final stops in dutch · the voicedness of intervocalic...

12
The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar stops in Dutch can be either voiced or voiceless, they are normally voiceless in word-final position. 1 This has been attributed to a rule of Final Devoicing. The rule is assumed to devoice either any obstruent at the end of a word (Trommelen and Zonneveld 1979), or any obstruent in coda position (Booij 1980). The conditions under which intervocalic word-final stops may be pronounced as voiced have been discussed in the phonological literature. Sometimes, word- final stops are voiced instead of voiceless when they are uttered postvocalically and before a vowel-initial function word. Thus, /heb Ik/ 'have V is sometimes pronounced with a [b] ([heb Ik]), although [hep Ik] is the expected pronunciation given the phonological rule of Final De voicing. The analyses given in the phonological literature are usually based on the intuitions and observations of the author. Although a good starting point, they might not accurately depict connected speech phenomena. Therefore, I con- structed a corpus of spontaneously spoken Dutch, which probably constitutes a more solid empirical base for the phonological investigation of connected speech. In this article, I will first discuss the corpus I built. I will argue that this corpus is suited to form the database for analyses of connected speech phenomena in Standard Dutch. Subsequently, the corpus is used for an investigation of the question under which conditions postvocalic word-final stops can be realized as voiced before vowel-initial clitics. I will consider the intervocalic alveolar and bilabial verb-form final stops followed by the function words ik [Ik], T , and het [at], 'it', with object function. It will be shown that the analyses given in the literature do not fit the data from my corpus. 1 I would like to thank Geert Booij, Vincent van Heuven, and Carlos Gussenhoven for their stimula- ting comments. Further, I am grateful to Marian Klamer, and Piet van Reenen for transcribing the stops. Finally, I would like to thank Evert Wattel for helping me with the statistics, Ton Wempe for helping me taping the speech for the corpus, and Audra Dainora for correcting my English in this paper. Linguistics in the Netherlands 1997, 7 3 – 4 . DOI 10.1075/avt.14.09ern ISSN 0929–7332 / E-ISSN 1569–9919 © Algemene Vereniging voor Taalwetenschap

Upload: trinhquynh

Post on 16-Aug-2019

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

The Voicedness of Intervocalic Word-Final Stops in Dutch

Mirjam T. C. Ernestus

0. Introduction

Although bilabial and alveolar stops in Dutch can be either voiced or voiceless, they are normally voiceless in word-final position.1 This has been attributed to a rule of Final Devoicing. The rule is assumed to devoice either any obstruent at the end of a word (Trommelen and Zonneveld 1979), or any obstruent in coda position (Booij 1980).

The conditions under which intervocalic word-final stops may be pronounced as voiced have been discussed in the phonological literature. Sometimes, word-final stops are voiced instead of voiceless when they are uttered postvocalically and before a vowel-initial function word. Thus, /heb Ik/ 'have V is sometimes pronounced with a [b] ([heb Ik]), although [hep Ik] is the expected pronunciation given the phonological rule of Final De voicing.

The analyses given in the phonological literature are usually based on the intuitions and observations of the author. Although a good starting point, they might not accurately depict connected speech phenomena. Therefore, I con­structed a corpus of spontaneously spoken Dutch, which probably constitutes a more solid empirical base for the phonological investigation of connected speech.

In this article, I will first discuss the corpus I built. I will argue that this corpus is suited to form the database for analyses of connected speech phenomena in Standard Dutch. Subsequently, the corpus is used for an investigation of the question under which conditions postvocalic word-final stops can be realized as voiced before vowel-initial clitics. I will consider the intervocalic alveolar and bilabial verb-form final stops followed by the function words ik [Ik], T , and het [at], 'it', with object function. It will be shown that the analyses given in the literature do not fit the data from my corpus.

1 I would like to thank Geert Booij, Vincent van Heuven, and Carlos Gussenhoven for their stimula­ting comments. Further, I am grateful to Marian Klamer, and Piet van Reenen for transcribing the stops. Finally, I would like to thank Evert Wattel for helping me with the statistics, Ton Wempe for helping me taping the speech for the corpus, and Audra Dainora for correcting my English in this paper.

Linguistics in the Netherlands 1997, 73–4. DOI 10.1075/avt.14.09ern ISSN 0929–7332 / E-ISSN 1569–9919 © Algemene Vereniging voor Taalwetenschap

Page 2: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

74 MIRJAM T.C. ERNESTUS

1. The corpus

The corpus consists of 7.5 hours of spontaneous speech produced by ten men, who were born and raised in the western part of the Netherlands. Each subject has an academic degree, speaks Dutch as his native language, and has no known speech impediments.

In a soundproof room at the phonetics laboratory of the University of Amster­dam, I recorded the men in groups of two. Each pair of men recorded together knew each other very well. Each subject had his own microphone, and the speech was recorded on different tracks of a DAT-tape.

First the men discussed a wide range of subjects for forty-five minutes. To ensure a steady stream of conversation I occasionally asked the men questions about where they had first met, their jobs and their hobbies. Some of the pairs talked for the entire period without interruptions and did not need prompting from me.

After forty-five minutes, I left the room. During the next three quarters of an hour the men had to play roles in a sociodrama. I hoped that while playing the drama, the subjects would forget that their speech was being taped. The socio­drama, in which one of the men visited the other one, consisted of three parts.

In the first part, the men were asked to talk about a party at which they discussed a broad range of topics with several people. Then they had to describe their visits to the dentist and their driving tests. This part lasted about fifteen minutes.

The second part of the sociodrama -- which was inspired by van der Wijst's (1996) negotiation game -- involved role-playing between a salesman and a camping store owner. The man acting as the salesman was instructed to sell tents, sleeping bags and backpacks. He was told that the goods came in packages of one-hundred only and that he could not sell the products separated. The man acting as the store owner was instructed to try to purchase a maximum of 75 sleeping-bags and 75 backpacks and not to buy tents. The salesman only knew the production costs of his products, while the shop owner only knew the prices of the products in his shop.

During the last part of the sociodrama, the men were instructed to pretend that they were having a drink and were talking about their sport-clubs. Most subjects did not act out this part of the sociodrama because they used up the entire forty-five minute period acting out the first two parts of the sociodrama.

Six Dutch male academics listened to the taped conversations and determined that the subjects were speaking (western) Standard Dutch. Three of the judges were born and raised in the western part of the Netherlands, while the other three were born and lived part of their lives in other parts of the country. The conver­sations sounded like natural spontaneous speech to the listeners. Some pairs of speakers spoke the most naturally when I was with them in the soundproof room, while the speech of other pairs was the most natural during the sociodrama.

Page 3: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

THE VOICEDNESS OF INTERVOCALIC WORD-FINAL STOPS IN DUTCH 75

It can be concluded that the corpus I built consists of (a western variant of) Standard Dutch. Since the speech it contains sounds natural and spontaneous, the corpus is suitable for analyses of connected speech phenomena.

2. The transcriptions

All combinations of a verb form ending in an alveolar or bilabial stop, and ik [Ik] T or het [ət] 'it' with object function were examined. A distinction was made between utterances with voiced intervocalic stops and utterances with voiceless intervocalic stops. I did this by listening to the strings containing the entire vowel preceding the relevant stop, the stop itself, and the entire vowel following it. I included the vowels in the strings since the perception of the voicedness of a stop is dependent on the formant transitions in the surrounding vowels and on the duration of the vowel preceding it (Slis 1985). The strings were isolated from the surrounding speech by means of the speech analysis software Praat (Boersma 1996).

I did not include in the database utterances in which the verb forms were separated from the following function words by glottal stops or hesitations, or utterances in which the stops were not followed by voiced vowels. I also did not include utterances which were repeated several times in succession or occurred in quoted speech. Repeated utterances may not be produced in the same way as utterances spoken only once, whereas quoted speech may not agree with the grammatical system of the speaker. When quoting, a speaker may be imitating someone else's grammatical system.

Two trained phoneticians judged the voicedness of the intervocalic word-final stops in the same way I did. A given stop was not included in the database if the judgment of both phoneticians did not agree with mine, or if the voicedness of the stop was judged as "unclassifiable" by at least one of us.

The data obtained in this way form the database for the investigations pres­ented in section 3. They can be found in Tables 1 and 2. The fourth and fifth columns of these tables are only relevant for the sections on frequency. The numbers given in the fourth column are not equal to the sum of the numbers given in the second and third columns, because the voicedness of the stops in many tokens could not be/was not determined.

3. Conditions for voiced word-final intervocalic stops

In this section, I will consider the voicedness of alveolar and bilabial, postvocalic, stops at the end of verb forms followed by the function words ik [Ik], T , or het [ət], 'it', with object function. I will investigate whether the surface voicedness of the stop is influenced by the underlying voicedness of the

Page 4: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

76 MIRJAM T.C. ERNESTUS

stop, by the frequency of occurrence of the verb form - function word sequence, and by the type of function word following the stop influence.

I will use Fisher's exact test to demonstrate the influence of these factors. Fisher's exact test indicates the chance that the p's of two binomial distributions are equal. The advantage of this test is that it can be used for all kinds of distributions. No assumptions have to be made about the distributions of the phenomena under investigation. Moreover, Fisher's exact test can be applied to small numbers. A disadvantage of Fisher's exact test is that it is rather insensi­tive.

Table 1. Number of tokens of combinations consisting of a verb form and "ik", pronounced with voiced and with voiceless stops in my corpus; plus the total number of these combinations in my corpus and in the Eindhoven corpus of spoken Dutch.

Combination My corpus Eindhoven corpus

with voi d stops with voiceless stops total total

heb ik 'have F "76 ~ _ 130 "48 begrijp ik 'understand I* 3 0 8 5

koop ik 'buy I' 3 0 3 1 snap ik 'understand I' 0 1 3 1

loop ik 'walk I' 1 0 2 1 riep ik 'called I' 0 1 1 0

stap ik 'step I' 1 0 1 0

weet ik 'know I' "27 3 44 67 moet ik 'must I' 28 7 57 33

had ik 'had I' 20 7 29 49

laat ik 'leave I' 3 0 11 12

zit ik 'sit I' 7 0 9 6

zat ik 'sat I' 2 0 3 6

deed ik 'did I' 1 2 4 1

bied ik 'offer I' 2 0 3 0

schat ik 'estimate I' 1 0 2 0

leid ik 'conduct I' 1 0 1 0

Page 5: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

THE VOICEDNESS OF INTERVOCALIC WORD-FINAL STOPS IN DUTCH 77

Table 2. Number of tokens of combinations consisting of a verb form and "het", pronounced with voiced and voiceless stops in my corpus; plus the total number of these combinations in my corpus and in the Eindhoven corpus of spoken Dutch. Combination My Corpus Eindhoven corpus

with voiced stops with voiceless stops total total

heb het 'have it' 4 2 20 2 begrijp het 'understand it' 1 0 4 0

snap het 'understand it' 0 1 2 0

weet het 'know(s) it' 8 3 19 1 moet het 'must it' 4 1 8 3

had het 'had it' 1 4 11 0

doet het 'does it' 1 1 7 1

deed het 'did it' 1 0 5 0

bied het 'offer it' 2 0 3 0

gaat het 'goes it' 1 0 2 0

haat het 'hate it' 0 1 1 0

laat het 'leave it' 1 0 1 0

spuit het 'spray it' 1 0 1 0

ziet het 'see it' 1 0 1 0

3.1. Underlying voicedness. Booij (1985:24) claims that an intervocalic word-final stop can only be pronounced as voiced if it is underlyingly voiced. An underlyingly voiced stop is pronounced as voiced when it forms a syllable with the following vowel-initial clitic in the lexicon. It then appears in onset position and final devoicing does not apply to it. It surfaces with its underlying [voice]-characteristics. According to Booij, in some dialects only stops in lexicalised combinations can surface as voiced (see below).

I investigated whether the surface voicedness of intervocalic word-final stops is influenced by their underlying voicedness in the corpus I collected. The results for the alveolar stops show that the data from the corpus do not support Booij's claim: underlyingly voiced as well as underlyingly voiceless alveolar stops were often pronounced as voiced in intervocalic word-final positions (see Table 3). There appears even to be no statistically significant difference in surface voicedness between underlyingly voiced and underlyingly voiceless alveolar word-final stops.

Page 6: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

78 MIRJAM T.C. ERNESTUS

It is not possible to investigate the influence of underlying voicedness on the surface voicedness of bilabial intervocalic word-final stops, since the corpus contains only heb as verb form ending in a vowel-[b] sequence. If the final stop of heb is pronounced as voiced more often than the final bilabial stops of other verb forms, this may be due not only to the underlying voicedness of the word-final stop of heb, but also to other characteristics of heb, such as its high fre­quency.

Table 3. Absolute (N) and relative (%) numbers of tokens with voiced and voiceless stops broken down by underlying voicedness and by type of function word. 'V' stands for 'vowel'.

combination type with voiced stops with voiceless stops

.. Vd] verb form ik

.. Vt] verb form ik

24 (73%) 68 (87%)

9 (27%) 10 (13%)

.. Vd] verb form h e t

.. Vt] verb form het

4 (50%) 17 (74%)

4 (50%) 6 (26%)

3.2. Frequency of occurrence. Booij (1985:24) states that, at least in some dialects, underlyingly voiced word-final, intervocalic, stops can surface as voiced only when they occur in lexicalised phrases. In lexicalised combinations, the clitic is added to the verb form in the lexicon. This implies that the verb form final stop is in onset position before Final Devoicing applies. It surfaces with its underlying [voice]-characteristics. In non-lexicalised combinations the clitic is added to the verb form in the post-lexical part of the grammar. The stop is devoiced as it is still in coda-position when Final Devoicing is applied.

Lexically stored phrases are normally assumed to be highly frequent (Stem-berger and MacWinney 1988). Therefore, Booij seems to predict that word-final stops surface as voiced only in highly frequent combinations.

I used the corpus to investigate the relationship between the voicedness of intervocalic word-final stops and the frequency of occurrence of the combination in which these stops occur. Bilabial and alveolar stops are considered separately for the following reason. If voicing of stops is an effect of articulatory reduction, different stops may be affected in different degrees, as their articulatory and acoustic characteristics differ. The combinations with ik will be separated from the combinations with het, as het is more clitic-like than ik. The pronoun het cannot be stressed and starts with a schwa, which is impossible for prosodically independent words in Dutch (Booij 1995). This might influence the voicedness of the preceding stops. Finally, given the result of section 3.1, the combinations with verb forms ending in underlyingly voiced stops and combinations with verb forms ending in underlyingly voiceless stops will be examined together.

Page 7: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

THE VOICEDNESS OF INTERVOCALIC WORD-FINAL STOPS IN DUTCH 79

Among the combinations consisting of a verb form ending in a vowel-stop sequence and ik, the following occur frequently: heb ik ('have I'), had ik ('had F), moet ik ('must F), and weet ik ('know F). I consider these combinations to be frequent, because in my corpus and in the pilot part of the Eindhoven corpus of spoken Dutch (Uit den Boogaart 1975) they occur far more often than other combinations consisting of a verb form ending in a vowel-stop sequence and some variant of ik (see Table 1). In my corpus, the combinations heb ik, had ik, moet ik and weet ik occur at least 29 times, whereas the other verb forms are followed by ik only 11 times or less. In the pilot part of the Eindhoven corpus of spoken Dutch, had ik, heb ik, moet ik, and weet ik occur at least 33 times. The other combinations occur maximally 12 times in the Eindhoven corpus.

As can be seen in Table 4, 96 % of the heb ik combinations were pronounced with [b]'s. The non-frequent combinations of verb forms ending in a vowel -bilabial stop sequence and ik were pronounced with [b]'s in 80% of the cases. Therefore, both frequent and non-frequent combinations appear to be often pronounced with voiced intervocalic bilabial word-final stops. According to Fisher's exact test, the data indicate that the chance that bilabial, intervocalic, verb form-final stops followed by ik are pronounced as voiced as often in frequent combinations as in non-frequent combinations is 9.5 %. My corpus thus shows a slight tendency for stops in frequent combinations to be more often pro­nounced as voiced than stops in non-frequent combinations.

Table 4. Absolute (N) and relative (%) numbers of tokens with voiced and voiceless stops broken down by frequency of occurrence and by type of function word. 'V stands for 'vowel'.

combination type with voiced stops with voiceless stops

( [ . . V { b , p }]verb form ik) frequent 76 (96%) 3 (4%)

( [ . . V ( b , p }]verb form ik) non-frequent 8 (80%) 2 (20%)

([••V{t,d}] verb form ik)frequent

75 (82%) 17 (18%)

( [ . . V { t , a } ] v e r b f o r m ik)non-frequent 17 (89%) 2 (11%)

([ V{b,p}] verb form het)frequent

4 (67 %) 2 (33 %)

([..V{b,p}]verb form het) non-frequent

1 (50%) 1 (50%)

([..V{t,d}] verb form het)frequent

13 (62%) 8 (38%)

( [ V { t , d } ] v e r b t o r m het)non-frequent 8 (80%) 2 (20%)

No statistically significant difference was found between frequent and non-frequent combinations of verb forms ending in /Vt/ or /Vd/, and ik. Both the alveolar stops in had ik, moet ik, and weet ik, and the alveolar stops in the non-frequent combinations were often pronounced as voiced.

Page 8: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

80 MIRJAM T.C. ERNESTUS

Among the combinations of verb forms ending in a post-vocalic stop and het, I consider the combinations had het ('had it'), heb het ('have it'), moet het ('must it'), and weet het ('know(s) it') as frequent. These 4 combinations occur at least 11 times in my corpus and in the pilot part of the Eindhoven corpus of spoken Dutch, while the other combinations are found only 8 times or less (see Table 2). The difference between frequent and non-frequent combinations is not as great for the verb forms followed by het as for the verb forms followed by ik.

As can be seen in Table 4, the frequency of occurrence does not seem to influence the voicedness of intervocalic bilabial stops at the end of verb forms followed by het. The heb het combinations were pronounced with [b]'s in 67 % of the cases. The other combinations were pronounced with [b]'s in 50 % of the tokens. This difference is not significant, according to Fisher's exact test. This lack of significance may be due to the small number (2) of non-frequent combina­tions considered.

A statistically significant frequency effect is not found for the voicedness of the alveolar postvocalic final stops of verb forms followed by het. The combina­tions had het, moet het, and weet het were pronounced with [d]'s in 62 % of the cases. The infrequent combinations were pronounced with [d]'s in 80 % of the tokens. The frequent combinations appear to be pronounced with voiced word-final stops somewhat less often than the non-frequent combinations. This is the opposite of what was expected.

It can be concluded that the data from my corpus do not support the hypoth­esis that combinations have to be frequent to be pronounced with voiced intervocalic word-final stops. In addition to frequently uttered combinations, many non-frequent combinations were found with voiced word-final stops.

The data from the corpus indicate that frequency of occurrence may influence the voicedness of the intervocalic stop in combinations consisting of a verb form ending in a postvocalic bilabial stop and ik. The frequent combination heb ik is somewhat more often pronounced with a [b] than the non-frequent combinations. The combination heb ik is much more frequent than all other combinations considered to be frequent in this paper. In the Eindhoven corpus and in my corpus, heb ik occurs 178 times in total. The combination weet ik, which is the second most frequent combination, occurs only 111 times in the corpora. It is possible that stops are pronounced as voiced more often only in highly frequent combinations, as only these combinations are lexicalised. Another possible explanation for the fact that heb ik seems to behave differently from the other combinations is that heb ik is frequent and its word-final stop is underlyingly voiced. This possibility is investigated in the next subsection.

3.3. Underlying voicedness and frequency of occurrence. In sections 3.1 and 3.2, the influences of underlying voicedness and frequency of occurrence on the

Page 9: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

THE VOICEDNESS OF INTERVOCALIC WORD-FINAL STOPS IN DUTCH 81

surface voicedness of stops in intervocalic word-final positions were investigated separately. It was found that these two factors did not influence voicing. It is possible that no influences were found because the two factors mask each other. The effect of underlying voicedness may only be found when the frequent and non-frequent combinations are separated from each other. Similarly, the influence of frequency of occurrence may only be found if the underlyingly voiced stops are separated from the underlyingly voiceless stops.

This hypothesis is not supported by the data from my corpus. First, no influence of underlying voicedness is found, when the frequent and the non-frequent combinations are separated. Second, no influence of frequency of occurrence is found when a distinction is made between combinations with underlyingly voiced word-final stops and combinations with underlyingly voice­less word-final stops. This is true for alveolar stops followed by ik and for alveolar stops followed by het (see Table 5). The hypothesis cannot be tested for bilabial stops, as pointed out in section 3.1.

Table 5. Absolute (N) and relative (%) numbers of tokens with voiced and voiceless stops broken down by underlying voicedness, by type of function word and by frequency of occurrence. 'V stands for 'vowel'.

Combination type with voiced stops with voiceless stops

( Vd] verb form ik )frequent 20 (74%) 7 (26%)

( Vt] verb form ik )frequent 55 (85%) 10 (15%)

non.-frequent 4 (67%) 2 (33%)

( Vd] verb form ik )non-frequent 13 (100%) 0 (0%)

( Vd] verb form het)frequent 1 (20%) 4 (80%)

( Vd] verb form het)frequent 12 (75%) 4 (25%)

-frequent 3 (100%) 0 (0%)

-frequent 5 (71%) 2 (29%)

3.4. Type of word following the stop. The frequency of intervocalic word-final stops being pronounced as voiced may depend on the word following the stop. Thus, according to Gussenhoven (1986:180), in Standard Dutch, bilabial and alveolar stops can only be voiced before ik. Berendsen (1986:55) claims that intervocalic, word-final /t/'s can only be voiced when followed by schwa-initial function words, such as het.

In my corpus, stops are frequently pronounced as voiced before ik as well as before het. This is true for both alveolar and bilabial stops (see Table 6).

However, alveolar stops preceding ik do seem to differ from stops preceding het in the way predicted by Gussenhoven. The data show a tendency for stops to

Page 10: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

82 MIRJAM T.C. ERNESTUS

be pronounced as voiced more often before ik than before het (p = 0.06). This trend is not present among the non-frequent combinations only.

Berendsen's claim that /t/'s can only be voiced in front of schwa-initial clitics does not hold for the data in my corpus. In 68 out of 78 cases, intervocalic /t/'s were pronounced as voiced in front of ik. Only in 2 of these combinations the vowel of ik was reduced to schwa.

Table 6. Absolute (N) and relative (%) numbers of tokens with voiced and voiceless stops broken down by type of function word for all combinations and for non-frequent combinations only (insignificance of the odds ratio deter­mined by Fisher's exact test).

Combination type with voiced stops with voiceless stops Chance that the 2 combination types do not differ

V{t,d}] verb form ik 92(83%) 19 (17%) 0.07

V{t,d}] verb form het 21 (68%) 10 (32%)

V{t,d}] verb form ik)non-frequent 17 (89%) 2(11%) > 0.1

V{t,d}] verb form het)non-frequent 8 (80%) 2 (20%)

V{p,d}] verb form ik 8 (80%) 2 (20%) > 0.1

V{t,d}] verb form het 5 (63%) 3 (37%)

3.5. Frequency of occurrence revisited. In sections 3.2 and 3.3 the influence of the frequency of occurrence of a combination on the voicedness of its intervocalic word-final stop was investigated. The only frequency effect found was a slight tendency to voice the intervocalic word-final stop in the highly frequent combina­tion heb ik more often than in the non-frequent combinations of verb forms ending in a postvocalic bilabial stop and ik. No difference in the voicedness of intervocalic word-final stops could be found between the frequent and non-frequent tokens of other kinds of combinations.

It may be that no influence of frequency was found because frequency was considered as an all-or-none property. Frequency was supposed to be a binary property, because it was used as an indication for lexicalisation: only frequent combinations may be lexicalised as a unit.

Perhaps, frequency of occurrence should not be considered to be a binary feature but a continuum. Bybee (1997), among others, argues for this. She states that 'sound change takes place in tiny increments in real time as words are used'. The more a word or phrase is used, and thus the higher its frequency of occur­rence, the more it is exposed to reduction. Frequency influences the realisation of a word, not as an all-or-none property, but as a continuum.

Page 11: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

THE VOICEDNESS OF INTERVOCALIC WORD-FINAL STOPS IN DUTCH 83

The hypothesis was tested that there exists a relation between voicedness and frequency of occurrence when frequency of occurrence is considered as a continuum. The correlation between the natural logarithm of a combination's frequency and its percentage of tokens with voiced word-final stops was calcu­lated with linear regression. The frequency of a combination was considered to be equal to its number of tokens in my corpus and in the pilot part of the Eind­hoven corpus of spoken Dutch (see Tables 1 and 2). The reliability of the percen­tage of tokens with voiced stops for each combination was considered to be equal to the natural logarithm of the total number of tokens with voiced and voiceless stops of that combination plus 1. In this way, the percentages of combinations with many tokens were weighted more heavily than the percentages of combina­tions with few tokens. This was done since the percentages for combinations with many tokens are more reliable.

If all combinations of verb forms ending in an alveolar or bilabial stop and ik or het are considered together, frequency of occurrence probably correlates with the percentages of tokens with voiced word-final stops. A correlation of 18 % is found (p < 7 %). If the combinations are broken down by place of articulation of the verb form final stop and/or by the function word, the unreliability of the correlation increases.

Note that an effect of frequency of occurrence is found with linear regression. Linear regression is preferably used with normal distributions. Kendall's Tau, which can be used with all kinds of distributions, does not show an effect of frequency.

4. Conclusions

My corpus contains more tokens of voiced, word-final inter-vocalic stops than could be expected given the Dutch phonological literature. First, both under-lyingly voiced and underlyingly voiceless stops are frequently pronounced as voiced. This implies that word-final stops do not surface as voiced only if they are underlyingly voiced and do not undergo Final Devoicing. Second, voiced word-final stops are often found in frequent as well as in non-frequent combina­tions. The voiced production of word-final stops appears to be not only a matter of lexicalisation. Third, many voiced post-vocalic stops are found before het, as well as before ik.

Although both frequent and non-frequent combinations can be pronounced with voiced intervocalic word-final stops, frequency of occurrence may influence the voicedness of these stops. The corpus shows a tendency for the intervocalic stops in the highly frequent combination heb ik to be pronounced as voiced more often than the verb form final bilabial stops in non-frequent combinations. Moreover, the corpus shows a possible correlation between the frequency of occurrence of a combination and its percentage of tokens with voiced stops, when

Page 12: The Voicedness of Intervocalic Word-Final Stops in Dutch · The Voicedness of Intervocalic Word-Final Stops in Dutch Mirjam T. C. Ernestus 0. Introduction Although bilabial and alveolar

84 MIRJAM T.C. ERNESTUS

frequency of occurrence is considered as a continuum and all combinations are considered together. Perhaps, the two attested trends are significant effects in larger corpora of spontaneous speech. This possibility is investigated in my ongoing research.

It can be concluded that the intuitions of the Dutch phonologists do not agree very well with the data from my corpus. There are several possible explanations. Perhaps, the phonologists were wrong or have based their analyses on other variants of standard Dutch than I did. It may also be possible that the voicedness of the stops in my corpus is not only determined by the phonology. Some stops may be perceived as voiced simply because they are very short. This is another possibility being investigated in my current research.

References

Berendsen, E. (1986) The Phonology of Cliticization, ICG Printing, Dordrecht. Boersma, P (1996) 'Praat, a System for Doing Phonetics by Computer', ms. University of Amster­

dam. Booij, G.E. (1980) 'De Syllabe in de Generatieve Fonologie', Spektator 9, 548-58. Booij, G.E. (1985) 'Lexical Phonology, Final Devoicing and Subject Pronouns in Dutch', in H.

Bennis and F. Beukema, eds., Linguistics in the Netherlands, Foris, Dordrecht. Booij, G.E. (1995) The Phonology of Dutch, Clarendon Press, Oxford. Bybee, J. (1997) 'Lexicalization of Sound Change and Alternating Environments', in J. Pierrehumbert

and M. Broe, Laboratory Phonology V, Cambridge University Press, Cambridge (in press). Gussenhoven, C. (1986) 'Over de Fonologie van Nederlandse Clitica', Spektator 15, 180-200. Slis, I. H. (1985) The Voiced-Voiceless Distinction and Assimilation of Voice in Dutch, PhD disserta­

tion Rijksuniversiteit Utrecht. Stemberger, J. P. and B. MacWinney (1988) 'Are Inflected Forms Stored in the Lexicon?', in M.

Hammond and M. Noonan, eds., Theoretical Morphology, Academic Press, San Diego. Trommelen, M. and W. Zonneveld (1979) Inleiding in de Generatieve Fonologie, Dick Coutinho,

Muiderberg. Uit den Boogaart, P.C., red. (1975) Woordfrequenties in Geschreven en Gesproken Nederlands,

Oosthoek, Scheltema and Holkema, Utrecht. Van der Wijst, P. (1996), Politeness in requests and negotiations, PhD dissertation, Katholieke

Universiteit Brabant, Tilburg.