• Dutch
  • Frisian
  • Saterfrisian
  • Afrikaans
Show all
Regressive voice assimilation

Regressive voice assimilation (RVA) is a well-known phonological process, extensively present in Afrikaans, much more than in Dutch.


In Dutch, RVA is restricted to 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; for extensive references, see the topic on RVA in Dutch. According to Le Roux and Pienaar (1927), for Afrikaans a similar situation existed until at least the early decades of the previous century i.e. only stop consonants acted as triggers for RVA. Wissing, in a number of publications (e.g. in Wissing, 1992), points out that since then this has changed dramatically – that is to say providing Le Roux and Pienaar's (1927) observation was correct. All voiced segments – fricatives, sonorant consonants, glides and even vowels to an extent, may cause RVA to take place. Moreover, RVA is present in different types of grammatical structures: compound words, derivations and in phrases, that is, both within words and across word boundaries.

As a point of departure (in Section A), RVA in Afrikaans compounds is described, taken from Wissing (1990) and, to an extent, from Combrink and De Stadler (1987). Subsequently, in Section B, results are reported of an experiment (see Wissing (1992)) in which the occurrence of RVA in compounds, derivations and in phrases was investigated. A preliminary scale is provided of the relative strength of voicing influence that segments have on preceding voiceless consonants.

In the final section C, the results provided in Sections A and B are compared to those obtained from an investigation of the presence of RVA in the RSG spoken database, mentioned in the introduction to phonological processes; we focus on readings by one male presenter of RSG news bulletins; these are characterised by the clear presence of regressive voice assimilation.

[+]A. Overview of RVA in Afrikaans compounds

Examples of RVA where obstruents are involved are as follows. (4.4) below serves as a representative for all types of fricatives combined in C1 and C2 positions i.e. /sv/, /fv/ and /xv/. In later sections, more instances of RVA, i.a. with sonorant consonants as triggering segments in C2-position, are provided. Following this list of examples, two representative examples of oscillograms and spectrograms, as well as accompaning sounds are given, one of a fricative, and another of a stop consonant as C1 – see Figure 1.

    1. /pd/ > [bd]: opdrag /ɔpdrɑx/ ['ɔbdrɑx] command
    2. /pb/ > [b]: rampbestuur /rɑmpbəstyr/ ['rɑmbəsty:r] disaster management
    1. /tb/ > [db]: kortbroek /kɔrtbruk/ ['kɔrdbruk] shorts
    2. /td/ > [d]: uitdien /œytdin/ ['œydin] serve
    1. /kb/ > [gb]: vakbond /fɑkbɔnd/ ['fɑgbɔnt] trade union
    2. -
    1. /sb/ > [zb]: busbestuurder /bœsbəstyrdər/ ['bœzbəsty:rdər] bus driver
    2. /fd/ > [vd]: hoofdirekteur /hofdirɛktør/ ['hovdirɛktør] chief director
    3. /xd/ > [ɣd]: lugdiens /lœxdins/ ['lœɣdins] airline
    4. /sv/ > [zv]: reddingswerkers /rɛdəŋsvɛrkərs/ ['rɛdəŋzværkərs] rescue workers

  1. In 1.2 a subsequent phonological process is common, viz. /pb/ > [bb] via RVA, and the /bb/ > [b] due to degemination (see Degemination). Mutatis mutandis the same applies to 2.2.
  2. 3.2 is empty; no relevant examples of words with the cluster /kg/ in compounds exist in Afrikaans, due to the presence of /g/ in loan words only.
  3. Products of RVA similar to these could be added should prefixes like op-, af-, uit- be considered as constituents of compounds. Examples are: opbreek break up, afbreek break down, and uitbreek break out. When RVA is applied here, the relevant clusters /pb/, /fb/ and /tb/ are realised as, respectively, [bb] (> [b]), [vb] and [db]. See below for an alternative treatment of forms similar to prefixes in derivations, as also provided in Wissing (1992).

Figure 1
[click image to enlarge]

Listen to the following example:

In Figure 1 vertical blue lines indicate the glottal pulses of voiced segments, clearly showing the regressive assimilative influence of following voiced [d] on /s/ and / [p/], rendering [z] and [b] respectively in is die and op die.

[+]B. Experimental results

Results regarding the productivity of the phonological process of RVA in Afrikaans compounds, derivations and phrases were provided in Wissing (1992). A summary of the findings follows. Based on the results, a scale of the relative strength of consonants with respect to RVA influence is proposed.


Compounds with the structure XC1#C2Y were read in carrier phrases by two Afrikaans speaking participants; it was known beforehand that their pronunciation was characterised by the presence of RVA. C1 could be either the plosive /k/ or the fricative /s/; the C2-position was occupied by segments representing plosives ( /b/, /d/ and /v/), fricatives ( /v/), the sonorant consonants ( /l/ (lateral) and /n/ (nasal)), approximants ( /j/ /ɦ/) and vowels ( /ɑ/ /a/). These segments were also utilised in derivations and phrases (see below). Compounds of the following type were thus formed: C1 = /s/(plaas farm + botter butter), and C1 = /k/ (vark pig + bakke bowl) rendering the lists:

  1. plaas- + -botter, - -dogter, -werker, -naam, -lewe, -jagter, -hekke, -akkers.
  2. vark- + -bakke, -dorings, -wagters; -niere, -lewer, -jagters, -hare, -are.

[+]RVA in derivations

Here the description is restricted to words starting with af-, uit-, as reported in Wissing (1992). Combinations with words starting with the voiced obstruent consonants /b/, /d/ and /v/, the sonorant consonants /n/ and /l/, as well as the approximants /j/, /ɦ/ and the vowel /ɑ/, rendered the following combinations:

  1. af- + -brekend, -dreigend, -wagtend, -nemend, -lopend, -jakkerig, -hangend, -ademend.
  2. uit- + -blakerig, -dagend, -wykend, -nemend, -lewerend, -jagerig, -houdbaar, -asem.

[+]In phrases

Two types of phrase were investigated, both in carrier sentences: phrases containing normal words and phrases containing surnames (indicated by X):

  1. Koos wou weet of X 'n Afrikaanse woord is. Koos wanted to know whether X is an Afrikaans word.
  2. Maar ek weet mos dat X 'n Afrikaanse woord is. But of course I know that X is an Afrikaans word.
  3. Ek dink dis X wat Piet wou spel. I think it is X that Piet wanted to spell

The focus words are: baar, daar, waar, naar, lelik, hopeloos, jammer, amper.

Inserted surnames are: Barendse, Daneel, Wouters, Nelson, Leenderts, Haarhoff, Albertyn.

Onsets of both types are respectively /b/, /d/, /v/, /n/, /l/, /ɦ/, /j/ and /ɑ/.

In cases of RVA the three consonants /f/ (in of (1)); /t/ (in dat (2)); and /s/ (in dis (3)) changed to the phones [v], [d] and [z] respectively.

The results of this investigation show that RVA is present in all instances, though not to the same degree. Obstruents are much stronger in influencing preceding segments' voice status, followed by sonorant consonants, approximants and to a very small degree also vowels. The RAP index for all the categories taken together (N = 651) is 0.44. Phrases with focus words inserted in the carrier sentence rendered a RAP of 0.64, while for phrases with surnames the RAP was 0.30. In comparison, compounds scored 0.23 while derivations scored 0.21. Clearly RVA is applied significantly more in phrases. Analysed per type of segment, obstruents are strongest (plosives 0.64; fricatives 0.62), followed by laterals (0.26) and nasals (0.24). The approximants scored a lowish 0.17, and vowels only 0.02. Based on these RAP indexes, a strength hierarchy was proposed by Wissing (1992).

Relative strength of C2-segments in RVA
Table 1: Relative strength of C2-segments in RVA
/b/ /d/ /v/ /l/ /n/ /ɦ/ /j/ /a/

This hierarchy is very much in line with the universally accepted Sonority Sequencing Principle proposed by Clements (1990). Statistical differences lead to a threefold distinction between plosives, fricatives and the rest of the segments.

[+]RVA in the RSG dataset of spoken Afrikaans

On the basis of the results of the foregoing experiment, the exploitation of the RSG dataset was confined to the presence of RVA in normal words in sentence context (notably in phrases). Furthermore, in order to limit the extent of this section, the same segment clusters as above, as read by one RSG-presenter, were investigated. Therefore, the focus is on the obstruents viz. the voiceless and voiced plosives and fricatives.

[+]RVA in phrases

This section constitutes an overview of the appearance of RVA in sentences in the RSG news bulletins as read by one RSG presenter, known to be a clear producer of RVA. The focus is on providing a representative sample in which obstruents are involved. Furthermore, for the purpose of comparison, the description is limited to the same clusters of consonants that are mentioned in (1) – (4) above, viz. /pd/, /pb/, /tb/, /td/, /kb/, /sb/, /fd/, /xd/ and /sv/.

    1. /pd/ > [bd]: op dieselfde /'ɔp di'sɛlfdə/ ['ɔb di'sælfdə] on the same
    2. /pb/ > [b]: op Bethlehem /ɔp bɛtliɦɛm/ [ɔ-'bɛtliɦɛm] in Bethlehem
    1. /tb/ > [db]: dit bring /dət brəŋ/ [dəd brəŋ] this brings
    2. /td/ > [d]: het die /ɦɛt di/ [ɦɛ- di] have the
    1. /kb/ > [gb]: week bekend /vek bəkɛnd/ [viəg bəkɛnt] week known
    1. /sb/ > [zb]: lopies byvoeg /lopis bəifux/ [luəpiz bəifux] adding on runs
    2. /fd/ > [vd]: of die /ɔf di/ [ɔv di] or the
    3. /xd/ > [ɣd]: Johannesburg dood /joɦɑnəsbœrx dod/ [joɦɑnəsbœrɣ duət] Johannesburg dead
    1. /sv/ > [zv]: dis waar /dəs var/ [dəz var] it's true

Figure 2: Rule Application Probabilities of Regressive Voice Assimilation by one radio presenter. RAPs are on top of the columns
[click image to enlarge]
  1. Large interindividual and even intraindividual variation exists in the total RSG spoken database; the results should be viewed in this light.
  2. RVA is markedly more present in the readings of the male presenters.
  3. Generally RVA is clearly present in the pronunciation of the RSG-presenter whose results are given above.
  4. Complete 100% assimilation is found in the two cases where homorganic plosives are involved, viz. in 1b and 1c.
  5. Even though the consonants /f/ and /d/ in 2b are not homorganic, RVA is present in all productions of this speaker.
  6. Two other plosive clusters (1a and 1d), with voiceless /p/ and /t/ as C1's, also have very high RAP scores ( > 0.80); /k/ is far less prone to RVA (RAP = 0.50 and 0.60), but there is nonetheless still clear evidence for the widely spread occurrence of this type of assimilation.
  7. A characteristic of Afrikaans RVA, different than in Dutch is the presence of RVA where C2 is a fricative, as is clearly seen in 3.
  8. The allophones [g] (in (1.e) and (1.f), [z] (in (2.a) and (3), as well as [ɣ] (in (2.c) are less common though nevertheless still above 0.50, demonstrating the prolific strength of RVA as a phonological process in Afrikaans.

In the word lists below more examples are provided of the ten types of phrasal combinations of obstruents as provided in the tables below.

RVA tipes
Table 2: RVA tipes
/pd/ > [bd] (1a) /pb/ > [b] (1b) /td/ > [d] (1c) /tb/ > [db] (1d) /kd/ > [gd] (1e) /kb/ > [gb] (1f)
daarop dat dorp beset Duitsland die besigheid by boek deur aansoek by
Kemp drie groep beswaardes gemeet deur front bibber Derick Duvenhage gemaak by
op dertien kusdorp by het die het benadruk maak daarop ongeluk betrokke
op dieselfde op buitelandse luitenant Derrick met boere ongeluk dood ook bekommerd
op drie stap besluit moontlikheid dat rand by ook drasties ook bevind
stormloop dood voedselhulp benodig sowat dertig strand buite tronk dood versoek beslis
waarop die vragskip by vanoggend deel wat bekend vermoedelik deur week bekend
Table 3
/sd/ > [zd] (2a) /fd/ > [vd] (2b) /xd/ > [ɣd] (2c) /sv/ > [zv] (3)
as dertig asof die berig dat as wraak
as die homself daarna klopjag deur burgers woon
gevrees dat hulleself die nog drie gewys word
is diep of daar Saterdag deur inwoners wat
ses-drie of dat slag dood is Woensdag
tydens die of dokumente verwag dat ooreenkoms wat
weens die of driehonderd volg die uiters warm
  1. RVA mostly but not exclusively occurs with function words like is, het and of in the affected position, but as is clear from the examples, content words are also influenced, e.g. nouns such as stormloop in (1.a), verbs like gemeet in (1.3), and adjectives such as uiters in (3).

  • Clements, George N1990The role of the sonority cycle in core syllabificationPapers in Laboratory Phonology1Cambridge University Press283-333
printreport errorcite