• Dutch1
  • Frisian
  • Saterfrisian
  • Afrikaans
Show all
Voice assimilation
quickinfo

Dutch presents a process of voice assimilation in obstruent clusters which are the result of compounding.

In consonant clusters consisting of two obstruents, the voicing of the two obstruents will always be identical in the surface form, also in cases where these obstruents have different phonological voicing specifications. Yet, depending on the manner of articulation of the obstruents, the outcomes differ: the consonant cluster will be voiced if the second consonant is a voiced plosive; in all other combinations, the resulting consonant cluster surfaces as voiceless. Consider the overview in table 1 below. /b/, /p/, /z/ and /s/ represent voiced or voiceless plosives or voiced or voiceless fricatives, respectively.

These patterns have been attributed to different phonological processes, in particular regressive voice assimilation (RVA) versus progressive voice assimilation/ fricative devoicing.

Table 1
1st obstruent 2nd obstruent Plosive Fricative
[+voice] [-voice] [+voice] [-voice]
(RVA) (Final devoicing) (Final devoicing) (Final devoicing)
Plosive [+voice] /bb/ [bb] /bp/ [pp] /bz/ [ps] /bs/ [ps]
[-voice] /pb/ [bb] /pp/ [pp] /pz/ [ps] /ps/ [ps]
Fricative [+voice] /zb/ [zb] /zp/ [sp] /zz/ [ss] /zs/ [ss]
[-voice] /sb/ [zb] /sp/ [sp] /sz/ [ss] /ss/ [ss]

Voicing that takes place at the phrase level will be described below in section Fricative Voicing and in the topic about casual speech.

extra
The feature ±voicein Dutch

In Dutch, there exists an opposition between voiced and voiceless plosives, and voiced and voiceless fricatives. The feature±voiceis thus distinctive in the class of obstruents. No opposition exists between voiced and voiceless sonorants: all sonorants are phonetically voiced. The feature ±voice is thus not distinctive in the class of sonorants.

Dutch has the following obstruents: /p/, /t/, /k/, /f/, /s/, /x/, /b/, /d/, /v/, /z/, /ɣ/, /h/. These can be subdivided according to voicing:

  • -voice: /p/, /t/, /k/, /f/, /s/, /x/ ,
  • +voice: /b/, /d/, /v/, /z/, /ɣ/, /h/ .

The glottal fricative /h/ has no voiceless counterpart; furthermore, the voiceless velar plosive /k/ has no voiced counterpart in the native vocabulary, as /g/ only occurs in loanwords:

i
goal /gol/ goal
oregano /ɔregano/ oregano

The following list shows minimal pairs for plosives and fricatives:

ii
paard /pard/ [part] horse vs. baard /bard/ [bart] beard
pak /pɑk/ package vs. bak /bɑk/ vat
pit /pɪt/ kernel vs. bit /bɪt/ bridle bit
paal /pal/ pole vs. baal /bal/ bale
tak /tɑk/ branch vs. dak /dɑk/ roof
tik /tɪk/ tap vs. dik /dɪk/ fat
top /tɔp/ peak vs. dop /dɔp/ button
taal /tal/ language vs. daal /dal/ decline
c /se/ <c> vs. zee /ze/ sea
faal /fal/ fail vs. vaal /val/ pale

For some Dutch dialects, amongst others Standard Dutch, the voicing contrast in fricatives appears to be less stable than the voicing contrast in plosives. Many speakers do not make a distinction between /x/ and /ɣ/, and between /f/ and /v/ in word-initial position. The distinction between /s/ and /z/ appears to be the most stable one. In cases of neutralization, both the voiced and the voiceless fricative are realised as voiceless (Collins and Mees 1981; Van de Velde et al. 1996; Kissine et al. 2003). (See also topics about velar fricatives /x/ and /ɣ/ and labiodental fricatives /f/ and /v/.)

Status of /g/ in Dutch
Whether the voiced velar plosive /g/ is a phoneme of Dutch is subject to some debate, as it only occurs in loanwords; however, there are some minimal pairs, e.g. the native word kool /kol/ cabbage versus the loanword goal /gol/ goal (cf. topic velar plosives /k/ and /g/).

Phonetics of the voicing contrast

Dutch is a so-called pre-voicing language: in word-initial positions, the voicing contrast in stops/plosives rests upon voice onset time (VOT; -70 ms for voiced, 20 ms for voiceless, cf. Lisker and Abramson 1964; Slis and Cohen 1969; -80 ms for voiced, 0-25 ms for voiceless, cf. Kager et al. 2007). So, Dutch distinguishes between phonetically voiced and phonetically voiceless plosives. In this respect, Dutch (together with Yiddish and Afrikaans) differs from the other Germanic languages, which are said to be aspiration languages, i.e. they have a contrast between phonetically voiceless unaspirated and phonetically voiceless aspirated obstruents (Kager et al. 2007; Van Alphen 2007; Kerkhoff 2007).

Dutch also differs from other Germanic languages in that the VOT distinction mentioned above does not hold for all obstruents but primarily for stops. In Dutch fricatives, the phonetic correlates of the contrast are somewhat different from those in plosives: while they are traditionally labelled as +voice, they do not always have phonetic voicing (Slis and Van Heugten 1989). Phonetic voicing thus is not a distinctive feature of phonologically +voice obstruents; yet, as reported in Slis and Cohen (1969), Slis and Van Heugten (1989) and Kissine et al. (2003), there is a significant difference in duration: voiceless (or fortis) fricatives have a significantly longer duration than voiced (or lenis) fricatives. None of the experiments, however, found significant duration differences for the velar fricatives /x/ and /ɣ/.

Distribution of voiced and voiceless fricatives

There are some restrictions on the distribution of voiced versus voiceless fricatives, which do not hold for voiced versus voiceless plosives. Voiced fricatives can only occur after A-class vowels, whereas voiceless fricatives can only occur after B-class vowels:

i
knuffel /knʏfəl/ hug but *kneufel /*knøfəl/
heuvel /høvəl/ hill but *huvvel /*hœvəl/

This restriction does not hold for plosives. Both voiced and voiceless plosives can occur after A-class vowels and B-class vowels:

ii
ratel /ratəl/ rattle
radar /radɑr/ radar
adder /ɑdər/ adder
otter /ɔtər/ otter

For more information see the topic on general co-occurrence restrictions in rhymes.

readmore
[+]Regressive voice assimilation (RVA)

Regressive voice assimilation (RVA) occurs in consonant clusters where the second obstruent is a voiced plosive. In such cases, the first obstruent of the cluster agrees in voicing with the voiced plosive. Thus, phonologically voiceless coda obstruents become voiced according to RVA, and counter to final devoicing. Some examples are provided below:

1
strop-das /strɔp.dɑs/ [ˈstrɔbdɑs] tie
zand-bak /zɑnd.bɑk/ [ˈzɑndbɑk] sandpit (cf. zand [zɑnt] )
lach-bui /lɑx.bœy/ [ˈlɑɣbœy] laughter
huis-baas /hœyz.baz/ [ˈhœyzbas] landlord (cf. huis [hœys] )

If the second plosive in the cluster is voiceless, the whole cluster always surfaces as voiceless:

2
ijk-punt /ɛik.pʏnt/ [ˈɛikpʏnt] checkpoint
bloed-proef /blud.pruv/ [ˈblutpruf] blood test
druk-fout /drʏk.fɑut/ [ˈdrʏkfɑut] misprint

From a phonological perspective, it cannot be determined whether this assimilation in voicing quality is a case of RVA, or whether the cluster-initial obstruent surfaces as voiceless due to final devoicing; both processes would lead to the same outcome.

RVA not only applies in complex words but also in underived ones. The effect is visible in loanword adaptation, where combinations of a voiceless obstruent and a voiced plosive are never found, even if these clusters are allowed in the source language:

3
Lesbos /lɛs.bɔs/ [ˈlɛzbɔs] Lesbos
asbest /ɑs.bɛst/ [ˈɑzbɛst] asbestos
anekdote /anɛk.dotə/ [anɛgˈdotə] anecdote

Sonorant consonants and vowels do not trigger regressive voice assimilation:

4
zandloper /zɑnd.lopər/ [ˈzɑntlopər] [*zɑndlopər] sand glass
goudader /ɣɑud.adər/ [ˈɣɑutadər] [*ɣɑudadər] gold vein

In compound words with three-obstruent clusters, consisting of a word-final plosive-fricative cluster and a word-initial voiced plosive, regressive voice assimilation applies as well. The complete plosive-fricative cluster is pronounced as voiced:

5
fietsbel /fits.bɛl/ [fi̬t̬sbɛl] bell
Piets deken /pits.dekən/ [pi̬t̬sdekən] national dollar

If the third consonant of the cluster is a voiceless plosive, the whole cluster will be produced without voicing (regardless of the underlying voice values of this plosive-fricative cluster):

6
Jaaps tunnel /japs.tʏnəl/ [japstʏnəl] Jaap's tunnel

In three-consonant clusters where the third consonant is a sonorant consonant, voice assimilation only applies partially: the plosive-fricative cluster carries some voicing. If the plosive-fricative cluster is followed by /h/ or a vowel (phonetically starting with [ʔ], a voiceless stop), voice assimilation does not apply (Jansen 2007).

Phonetic or phonological?
Ernestus (2000) considers RVA to be a phonetic process rather than a phonological process. She states that obstruents in final position do not have phonological voice specifications. The phonetic values for voicing of these obstruents depend purely on the following segment. A voiced stop following the unspecified final obstruent is coarticulated together with this preceding final stop. In other words, voicing of the second obstruent starts during the first, unspecified, obstruent. The latter thus receives phonetic voicing from the former. If the final obstruent is followed by a phonologically voiceless obstruent, voicing cannot be coarticulated. The complete cluster thus surfaces as phonetically voiceless. Final obstruents followed by vowel-initial content words are phonetically voiceless (at least partially), since the vowels of these words are preceded by a glottal stop, which has no phonetic voicing. Therefore, the final obstruent surfaces as phonetically voiceless. Final obstruents preceding sonorant consonants remain subject to investigation. Phonetically they are voiceless, but it remains unclear how phonetics proper can account for this, since sonorants are phonetically voiced.

A final obstruent followed by a major phonological boundary (e.g. intonational phrase) is usually perceived as voiceless. Obstruents followed by a phonological boundary are ‘acoustically relatively long’, and long obstruents ‘tend to be perceived as voiceless’.

[+]Progressive Voice Assimilation or fricative devoicing

In obstruent-fricative clusters in compounds, the voicing of the two members of the consonant cluster is identical in the surface form: such clusters always surface as voiceless, regardless of the underlying voice values of the two members; thus even when both consonants are underlyingly voiced:

7
a. [+Voice] + [+Voice]
      huisvuil /hœyz.vœyl/ [ˈhœysfœyl] garbage
      rondvaart /rɔnd.vart/ [ˈrɔntfart] cruise
b. [+Voice] + [-Voice]
      zorgsector /zɔrɣ.sɛk.tɔr/ [ˈzɔrxsɛktɔr] healthcare
      ribfluweel /rɪb.fly.ʋel/ [ˈrɪpflyʋel] corduroy
c. [-Voice] + [+Voice]
      grijpgraag /ɣrɛip.ɣraɣ/ [ˈɣrɛipxrax] greedy
      drijfzand /drɛiv.zɑnd/ [ˈdrɛifsɑnt] quicksand
d. [-Voice] + [-Voice]
      lachstuip /lɑx.stœyp/ [ˈlɑxstœyp] fit of laughter
      drukfout /drʏk.fɑut/ [ˈdrʏkfɑut] misprint

As can be deduced from the these patterns, a fricative in the second position of such a cluster always surfaces as voiceless, regardless of its underlying specification for voice. This process is known as fricative devoicing, or progressive voice assimilation. The obstruent preceding the fricative always surfaces as voiceless as well, which can be attributed to final devoicing. From a phonological perspective, voicing of the following voiced fricative does not spread to the preceding obstruent, rather, it is -voiceof the preceding obstruent that spreads to the fricative.

Phonetic or phonological
Ernestus (2000) argues that initial fricatives must be specified for their underlying [voice]-features, since their realisation with regard to voicing is unpredictable in utterance-initial position. Fricatives preceded by another obstruent in clusters are always realised as voiceless since voicing is difficult to maintain in longer clusters: in fricative-fricative clusters, there is a relatively long period of friction, which inhibits glottal vibration (voice). In plosive-fricative clusters, the period of constriction followed by the period of friction inhibits voicing as well.

[+]Voicing in past tense suffixes

The Dutch past tense suffix surfaces as /-tə/ or /-də/ for the simple past, depending on the voice values of the final phoneme of the verb stem:

Table 2
Final segment (underlying) Past tense suffix
Vowel /-də/
Sonorant consonant /-də/
Voiced obstruent /-də/
Voiceless obstruent /-tə/

A similar pattern holds for the distribution of the present perfect and past perfect circumfix /ɣə-X-t/ or /ɣə-X-d/.

The Dutch past tense suffix surfaces as /-tə/ or /-də/ for the simple past, and as the circumfix /ɣə-X-t/ or /ɣə-X-d/ for the present perfect and past perfect. The choice of which suffix depends on the voice values of the final phoneme of the verb stem. If the verb stem ends in a vowel or sonorant consonant, the simple past suffix will be /-də/ (Zonneveld 1983,2007; Booij 1995; Grijzenhout and Krämer 1999; Ernestus 2000):

8
noemde /numdə/ called (inf. noemen )
roerde /rurdə/ stirred (inf. roeren )
voelde /vuldə/ felt (inf. voelen )
aaide /ajdə/ stroked (inf. aaien )
skide /skidə/ skied (inf. skiën )

If the final phoneme of the verb stem is an obstruent, the choice of /-tə/ or /-də/ depends on the (underlying) voice specification of the stem-final obstruent: the suffix will be /-tə/ if the stem-final obstruent is underlyingly voiceless, and it will be /-də/ if the stem-final obstruent is underlyingly voiced:

9
klapte /klɑptə/ clapped (inf. klappen )
lette /lɛtə/ watched (inf. (op)letten )
jankte /jɑŋktə/ cried (inf. janken )
fietste /fitstə/ cycled (inf. fietsen )
mafte /mɑftə/ snoozed (inf. maffen )
juichte /jœyxtə/ cheered (inf. juichen )
10
krabde /krɑbdə/ scratched (inf. krabben )
redde /rɛdə/ saved (inf. redden )
reisde /rɛizdə/ travelled (inf. reizen )
proefde /pruvdə/ tasted (inf. proeven )
legde /lɛɣdə/ layed (inf. leggen )

The choice of which suffix to select in the case of the present perfect or past perfect circumfix is made in the same way; however, this is only visible if the verb is adjectivized and followed by a schwa. If used as a verb or in attributive position without a final schwa, the final consonant will always be voiceless: the final /d/ of the circumfix undergoes final devoicing, and thus surfaces as [t]:

11
genoemd /ɣənumd/ [ɣənumt] called (inf. noemen )
geroerd /ɣərurd/ [ɣərurt] stirred (inf. roeren )
gevoeld /ɣəvuld/ [ɣəvult] felt (inf. voelen )
geaaid /ɣəajd/ [ɣəajt] stroked (inf. aaien )
geskid /ɣəskid/ [ɣəskit] skied (inf. skiën )
12
geklapt /ɣəklɑpt/ clapped (inf. klappen )
gelet /ɣəlɛt/ watched (inf. letten )
gejankt /ɣəjɑŋkt/ cried (inf. janken )
gefietst /ɣəfitst/ cycled (inf. fietsen )
gemaft /ɣəmɑft/ snoozed (inf. maffen )
gejuicht /ɣəjœyxt/ cheered (inf. juichen )
13
gekrabd /ɣəkrɑbd/ [ɣəkrɑpt] scratched (inf. krabben )
gered /ɣərɛd/ [ɣərɛt] saved (inf. redden )
gereisd /ɣərɛizd/ [ɣərɛist] travelled (inf. reizen )
geproefd /ɣəpruvd/ [ɣəpruft] tasted (inf. proeven )
gelegd /ɣəlɛɣd/ [ɣəlɛxt] layed (inf. leggen )

Phonologically, the circumfix must have a /d/ rather than a /t/, which can be seen if the past participle is used as an adjective (Zonneveld 2007):

14
genoemde /ɣənumdə/ [ɣənumdə] called
geroerde /ɣərurdə/ [ɣərurdə] stirred
gereisde /ɣərɛizdə/ [ɣərɛizdə] travelled
[+]Fricative voicing

Fricative voicing is a phonological process at the phrase level: voiceless fricatives that appear at the end of a prosodic word can surface as voiced if they are preceded by a sonorant consonant or vowel, and followed by a vocoid in the following prosodic word (Booij 1995):

15
pas op /pɑs-ɔp/ [pɑzɔp] watch out
twaalf uur /twalf-yr/ [twalvyr] twelve o’clock
was je /wɑs-jə/ [wɑzjə] were you
wasje /wɑsjə/ [wɑsjə] some laundry

The examples was je and wasje indicate that the process only applies at the sentence level, i.e. in the two prosodic wordswas je; within the single prosodic word wasje, the rule does not apply.

In intervocalic position, the voiceless fricatives /f/ and /s/ may be realised as voiced (Collins and Mees 1981); this happens much less frequently with the velar fricative /x/(Mees and Collins 1982):

16
af en toe /ɑf-ən-tu/ [ɑfəntu] [ɑvəntu] occasionally
rasecht /rɑs-ɛxt/ [rɑsɛxt] [rɑzɛxt] pure-bred

More information about phonological processes at the phrase level can be found in the topic about phonological processes in casual speech.

[+]Voicing in clitics

Clitics in Dutch do not always behave in the same way with respect to voicing. There is a variety of phenomena discussed below; many of the processes are optional.

First, there is a group of clitics that can trigger both regressive voice assimilation and progressive voice assimilation. Van Haeringen (1955) was one of the first to attempt to describe this phenomenon. He argues that the group of words below which can form clitics through attaching to a preceding or following word, triggers progressive voice assimilation, while /d/-initial words usually trigger regressive voice assimilation when preceded by an obstruent:

17
de /də/ the
dit /dɪt/ this
deze /dezə/ this, these
dat /dɑt/ that
die /di/ that, which
daar /dar/ there
d’r /dər/ there, her
dan /dɑn/ than, then
dus /dʏs/ thus

Zonneveld (1983), Berendsen (1986), Booij (1995) and Ernestus (2000) noticed that these clitics do not only trigger progressive voice assimilation, but also regressive voice assimilation (see sections above). When triggering regressive voice assimilation, the final obstruent of the preceding word surfaces as voiced, even if it is underlyingly voiceless. When triggering progressive voice assimilation, the initial obstruent of the second word, i.e. the clitic, surfaces as voiceless, regardless of the underlying voice values of the preceding obstruent:

18
lees de /lez-də/ [lezdə] [lestə] read the
lees dit /lez-dɪt/ [lezdɪt] [lesdɪt] read this
lees deze /lez-dezə/ [lezdezə] [lestesə] read these
lees dat /lez-dɑt/ [lezdɑt] [lestɑt] read that
op die /ɔp-di/ [ɔbdi] [ɔpti] on this
heb daar /hɛb-dar/ [hɛbdar] [hɛptar] have there
geef d’r /ɣev-dər/ [ɣevdər] [ɣeftər] give her
dat dan /dɑt-dɑn/ [dɑdɑn] [dɑtɑn] that then
had dus /hɑd-dʏs/ [hɑdʏs] [hɑtʏs] had thus

According to Berendsen (1986) and Booij (1995), the clitic /də/ always surfaces as [tə] when it follows a final /t/:

19
hij leest de krant [lest-tə]   → [lestə] [*lesd.də]  → [*lezdə] he reads the newspaper

The clitics -ie and -er (variants of die and der) always surface after /t/- and /d/-final words. Die and der can never surface in these contexts:

20
stond 'ie [stɔnti] [*stɔndi] stood he
vond d'r [vɔntər] [*vɔndər] found her

The other consonant-initial clitics of Dutch are only discussed by Berendsen (1986). The obstruent-initial clitics of Dutch are the clitics starting with /d/ (discussed above) and the clitics starting with /z/: -ze (subject use she, or object use them), -zen ( his) and -zich ( him/herself). When these clitics follow an obstruent-final word, they surface with initial [s] (cf. fricative devoicing). When they follow a sonorant consonant or a vowel, they remain /z/(Berendsen 1986):

21
gaf ze /ɣɑv-zə/ [ɣɑfsə] gave she/ them
had ze /hɑd-zə/ [hɑtsə] had she/ them
viel ze /vil-zə/ [vilzə] fell she
zou ze /zau-zə/ [zauzə] would she

The sonorant-initial clitics of Dutch are -je ( you), -we ( we), -me (<