• Dutch
  • Frisian
  • Saterfrisian
  • Afrikaans
Show all
Is the dorsal nasal /ŋ/ a cluster underlyingly?
quickinfo

This section deals with the question as to whether or not /ŋ/ is to be considered as the cluster /nG/ in underlying representation. The latter assumption seems to give a straightforward explanation for the distribution of [ŋ]. But there is also evidence which points in a different direction. All evidence taken together, it is concluded that the dorsal nasal is a single segment, also in underlying representation.

readmore

In the Dutch phonological literature, there was a debate concerning the underlying form of the dorsal nasal (see chapter 5 of Trommelen (1984) for a nice overview of the discussion and the arguments for and against). Some phonologists argued that Dutch [ŋ] derives from the underlying cluster /nG/ (where 'G' denotes some dorsal obstruent), since this would give a straightforward explanation for the distribution of [ŋ].

First, [ŋ] cannot occur in the syllable onset, because /nG-/ violates the Sonority Sequencing Principle (see: consonant sequences in general), according to which the onset must have a rising sonority profile.

Second, [ŋ] cannot follow a sequence of a short monophthong + a sonorant consonant or a long vocalic sequence, since this yields a syllable rhyme of four positions, viz. -/V{V/C}nG/, while the regular (word-final) rhyme can be occupied by three segments at most, as expressed by the Word Constraint (see the size of the word-internal and word-final syllable rhyme). These two arguments above arguably also hold for Frisian.

As to Frisian, there seems to be a third argument for adopting the cluster analysis. See the following examples of verb-diminutive pairs:

Table 1
Verbs Diminutives
ringje /rɪŋ+jə/ [rɪŋjə] to ring rinkje /rɪŋ+jə/ [rɪŋkjə] little ring
jongje /joŋ+jə/ [joŋjə] to drop young jonkje /joŋ+jə/ [joŋkjə] rummer of young gin
The Frisian diminutive morpheme has three allomorphs, viz. -ke, -tsje, and -je (see diminutive formation). The -je allomorph attaches to nouns ending in a velar consonant: so the diminutives of pak /pak/ parcel and skoech /sku:ɣ/ shoe are pakje [pakjə] and skoechje [skuxjə], respectively. When a noun ending in /-ŋ/ is diminuated, /k/ is inserted between the stem and the suffix (see rinkje and jonkje in the table above). Now, assuming [ŋ] to be /nk/ underlyingly would render /k/-insertion superfluous. However, in order to derive the verb ringje, the final /-k/ of the stem ring /rɪnk/ would have to be deleted, which is a complicating factor for the cluster analysis.

And there are more objections to this analysis. In the first place, the exact nature of the dorsal obstruent in the cluster /nG/ is unclear. It is not likely to be a fricative, i.e. /ɣ/ or /x/, since there is a strong cross-linguistic tendency for nasal consonants to assimilate to plosives, but not to fricatives (see Padgett (1994)). And indeed, [-cont] segments induce vowel nasalization in the the sequence vowel + /n/ by which they are preceded (see vowel nasalization).

This leaves the dorsal plosives /ɡ/ and /k/. The former only occurs in syllable-initial position, so it can be excluded right away (see the obstruents: the plosives). The latter also occurs in the cluster /-ŋk/ (see word-final clusters of a nasal and an obstruent). This implies that a cluster analysis would force us to assume stems ending in both /ŋk/ and /nk/. After assimilation of /n/ to /k/, both stems end in the same sequence, viz. /-ŋk/. However, /k/ in /ŋk/ derived from /nk/ must delete after assimilation, since only [ŋ] remains. This means either that /k/-deletion must be assigned the power to look back into the derivation in order to see whether or not preceding /ŋ/ derives from /n/ or that stems ending in /nk/ must be explicitly marked for undergoing /k/-deletion. Both options are undesirable.

Secondly, the cluster analysis does not result in a simpler lexicon, i.e. it does not enable us to restrict the underlying inventory of nasals in Frisian to just two, viz. labial /m/ and coronal /n/. The vowel /ɔ/ meets with the restriction that it can co-occur with neither /m/ nor /ŋ/ (see word-final single consonants). The same restriction holds when /m/ or /ŋ/ are the first member of the (homorganic) nasal + obstruent clusters /-m{p/b}/ and /-ŋk/, while /ɔ/ is allowed to combine with /p/ and /k/, as shown by kop /kɔp/ head and lok /lɔk/ lock, tress. This leads one to posit the following morpheme structure constraint: /ɔ/ + Nasal constraint, /*ɔ{m/ŋ}/.

This constraint implies that /m/ and /ŋ/ must be present as such in underlying representations, which means that, in view of lexical economy, there is nothing to be gained by deriving [-ŋ] from /-nk/. Again, in their surface forms the sequences [ɔm] and [ɔŋ] do occur, due to Regressive Place Assimiliation (see regressive place assimilation (nasal assimilation)), as in fan boppen /fɔn bopən/ [fɔm bopm̩] from the top, from above and (dat ik) de man ken /mɔn kɛn/ [mɔŋ kɛn] (that I) know the man.

Thirdly, like all coronal (dental-alveolar) consonants in Frisian, /n/ cannot be preceded by /a/, whereas it can by /ɔ/ (see word-final single consonants). This means that the cluster analysis forces one to derive the surface sequence [-aŋ] from /-ɔnk/, which necessitates lowering (and unrounding) of /ɔ/ to [a], which lacks independent phonological motivation. Moreover, the cluster analysis does not result in a simpler lexicon, since /a/ is needed for independent reasons. All evidence taken together, the dorsal nasal /ŋ/ must be assumed to be an underlying segment in Frisian.

As noted above, /ŋ/ has the distributional properties that it cannot occur in the syllable onset and neither after a sequence of a short monophthong + a sonorant consonant nor a long vocalic sequence. Trommelen (1984:177) calls these the cluster-properties of ŋ. They are accounted for by the Dorsal Nasal Constraint (see: the dorsal nasal /ŋ/), which exactly specifies in which position /ŋ/ occurs.

References
  • Padgett, J1994Stricture and Nasal Place AssimilationNatural Language en Linguistic Theory12465-513
  • Trommelen, Mieke1984The Syllable in DutchDordrechtForis
  • Trommelen, Mieke1984The Syllable in DutchDordrechtForis
printreport errorcite