Analysis of Assimilation in Dutch

Dutch has many examples of assimilation. Without question, assimilation is an important feature of the language. Two types of assimilation can be identified: progressive and regressive.

In the word zakdoek (handkerchief) the /k/ becomes /g/. This is an example of regressive assimilation. However, in the word drijfzand (quicksand) the /z/ becomes /s/. This is an example of progressive assimilation. Dutch also has a word-final devoicing rule which means that a word such as hond (dog) is pronounced with a word-final /t/. However, in compounds with a combination of a voiceless plosive and a voiced plosive, the voiceless plosive becomes voiced as in zakdoek. When a voiceless fricative is followed by a voiced one, the voiced fricative becomes voiceless as in drijfzand.

With certain cases, however, two analyses are possible. For example, zandbak (sandpit) is pronounced with a /d/. This can be analyzed as regressive assimilation. The reason is that zand (sand) is pronounced with a word-final /t/. This is the result of word-final devoicing.  The underlying representation is /zand/ and the phonetic representation is [zant]. The process can be illustrated as follows:

UR /zand/ + /bak/
word-final devoicing /zantbak/
PR regressive assimilation [zandbak]

However, it is also possible to argue that the voiced /d/ blocks word-final devoicing. If this is the case, the underlying representation is the same as the phonetic representation. This can be illustrated as follows:

UR /zandbak/
PR [zandbak]

Another example is the word bloedproef (blood test). In this compound the /d/ becomes /t/. This can be analyzed as a case of regressive assimilation. The process can be illustrated as follows:

UR /bloedproef/
PR Regressive Assimilation [bloetproef]

However, we can also argue that the surface form is the result of word-final devoicing. The voiceless /p/ fails to block the process. If this the case, the process can be illustrated as follows:

UR /bloedproef/
word-final devoicing /bloet/ + /proef/
PR [bloetproef]

Both analyses are plausible. However, phonologists tend to prefer the most elegant solution. In the word zandbak, the most elegant solution is to say the underlying representation surfaces because the voiced plosive of bak prevents word-final devoicing. This requires fewer rules than the other analysis. If we accept this analysis, we must say that the /t/ of bloedproef is the result of regressive analysis and not word-final devoicing.