• Dutch
  • Frisian
  • Afrikaans
Show full table of contents
Regressive Voice Assimilation of obstruent sequences

Regressive voice assimilation can affect a sequence of two or three obstruents (see examples of voice assimilation of fricative sequences triggered by non-plosive voiced segments and assimilation of obstruents sequences induced by a voiced plosive). This seems the be at odds with the general demand that phonological operations must be formulated as local processes. How the assimilation of obstruent sequences proceeds is the subject of this topic.


Regressive voice assimilation (RVA) can affect a sequence of two or three obstruents. If it is to be seen as a local phonological process, one cannot but assume that the obstruents in a sequence do not have their own, separate voice specifications, but that they, instead, share one and the same specification, so that the sequence as a whole can be voiced and devoiced in one swoop. Not only does this need to hold of underlying sequences, but also of the ones resulting from inflection, as those in (1):

Example 1

Examples of obstruent sequences resulting from inflection
(hy) bakt /bak+t/ (he) bakes
(do) skepst /skɛp+st/ (you) scoop, shovel
(wat) dreechs /dre:ɣ+s/ (something) difficult, hard
(wat) geks /ɡɛk+s/ (something) stupid, foolish

Such sequences are assimilated in one go as well.

Apart from the phonological plausibility, there is an independent piece of evidence for this analysis. Word-final sequences of a liquid and a plosive can be broken up by schwa epenthesis (see schwa insertion in word-final sequences). Word-final obstruent sequences, on the other hand, cannot, as shown in (2):

Example 2

Examples of the non-occurrence of schwa epenthesis in word-final obstruent sequences
krêft /krɛ:ft/ [krɛ:ft] [*krɛ:fət] strenght; power(s)
rjocht /rjoxt/ [rjoxt] [*rjoxət] justice; straight
rjochts- /rjoxts/ [rjoxts] [*rjoxəts] [*rjoxtəs] right(-hand)
lofts /lofts/ [lofts] [*lofəts] [*loftəs] left(-hand)
takst /takst/ [takst] [*taksət] [*takəst] regular amount
(hy) bakt /bak+t/ [bakt] [*bakət] (he) bakes
(wat) dreechs /dre:ɣ+s/ [dre:xs] [*dre:ɣəs] (something) difficult, hard

The shared voice specification of the obstruents cannot be interrupted by the independent voice specification of a vowel. A general constraint on obstruent sequences is that the obstruents must agree in voicing (see onset: sequences of two obstruents); this can be seen as the automatic consequence of their sharing one and the same voice specification.

[hide extra information]

Inflection often gives rise to word-internal sequences of obstruents with opposite voice specifications, as in (3):

Example 3

Examples of obstruent sequences with opposite voice specifications resulting from inflection
hy draacht /dra:ɣ+t/ [ dra:xt] (he) carries
(do) graafst /ɡra:v+st/ [ɡra:fst] (you) dig
(hy hat) wjud /vjød+d/ [vjøt < vjøtt < vjødt] (he has) weeded

Such sequences, however, end up as impenetrable: [*dra:ɣət] (draacht), [*ɡra:vəst] (graafst), [vjødət] (wjud), so they share their voice specification. This means that there must be a process of word-internal obstruent adaptation. A word-final obstruent is realized as voiceless, which implies that these sequences as a whole cannot but end up as voiceless as well (see final devoicing: obstruent sequences).

[hide extra information]

Like regressive place assimilation, regressive voice assimilation is also found in those acronyms each letter of which is pronounced separately, like FNP /ɛfɛnpe:/ [ɛvɛm'pe] name of a Frisian political party (i.e. de Fryske Nasjonale Partij the Frisian National Party), FA /ɛfa:/ [ɛv'ʔa:] (Fryske Akademy Frisian Academy), and FSW /ɛfɛsve:/ [ɛvɛz've:] name of one of the departments of the Fryske Akademy (i.e. the Fakgroep Sosjale Wittenskippen Department of Social Sciences). See also the extra on acronyms and regressive place assimilation.

[hide extra information]

Regressive voice assimilation regularly affects noun-final fricatives, as the following examples show:

Example 4

Examples of RVA affecting noun-final fricatives
Hast dy les al ynhelle? /lɛs ɔl/ [lɛz ʔɔl] have you already made up for that lesson?
Ik trof ien /trof iən/ [trov ʔiən] I met someone

Riemersma (1979:6) notes that there is no assimilation in case the fricative precedes a final schwa, which often deletes in connected speech (see schwa deletion as a synchronic proecess: how to deal with hiatus in syntactic configurations); see the following examples:

Example 5

Examples of the non-application of RVA with schwa-final nouns
Hast dy flesse al omspield /flɛsə ɔl/ [flɛs ʔɔl] [*flɛz ʔɔl] Have you already rinsed out that bottle?
Wy treffe ien /trɛfə iən/ [trɛf ʔiən] [trɛv ʔiən] We meet someone

In flesse al and treffe ien there is a configuration of vocalic hiatus. This is less favoured in Frisian, though between words it is not forbidden (see the resolution of vocalic hiatus in general). It can be remedied by the deletion of schwa. In order to explain the differences in assimilation between (4) and (5) then it might be argued that assimilation takes place before schwa deletion. The presence of schwa renders the fricative and the vowel/diphthong non-adjacent, wich blocks assimilation. Schwa deletion is an optional and highly variable process, whereas assimilation is categorical. One therefore expects the latter to precede the former.

Another explanation, however, is possible, and feasible. The facts of (5) can also be taken as indicative of the nature of schwa deletion in connected speech. Schwa may not really have been deleted here. The speaker simply does not manage to convert the phonemes of the words of these sentences in a one-to-one fashion into concrete speech sounds or, put differently, to interpret or to implement schwa in a proper way. At the phonological level, at which assimilation takes place, schwa remains in place, so that assimilation cannot apply.

[hide extra information]

There is no regressive voice assimilation in the names in (6) (Riemersma (1979:62)):

Example 6

Names in which RVA does not apply
Fryslân [frislɔ:n] Friesland
Yslân [islɔ:n] Iceland
Ruslân [røslɔ:n] Russia

RVA, however, is to be expected here, as shown by the examples in (7):

Example 7

Examples of compounds with -lân 'land' in which RVA applies
ryslân /ris#lɔ:n/ [rizlɔ:n] twig land
nimmenslân /nɪməns#lɔ:n/ [nɪmm̩zlɔ:n] no man's land
thúslân /tys#lɔ:n/ [tyzlɔ:n] homeland

The part -lân of the names in (6) thus no longer seems to function as an independent word − which it does in the examples in (7) −, but as a suffix, notwithstanding its full vowel.

It should be noted that Fryslân, Yslân, and Ruslân have a short vowel, which favours the voicelessness of the fricative (see the obstruents: the fricatives). The name of the Dutch province of Groningen, Grinslân /ɡre:nzlɔ:n/, has both a long (nasal) vowel and the voiced fricative [z]: [ɡrẽ:zlɔ:n].

  • Riemersma, Tr1979Sylabysjerring, nazzeljerring, assymyljerringLjouwertKoperative Utjowerij
  • Riemersma, Tr1979Sylabysjerring, nazzeljerring, assymyljerringLjouwertKoperative Utjowerij