Íslenskt mál og almenn málfræði - 01.01.1982, Page 311

Ritdómar
309
have, in accordance with Árnason's algorithm, a long vowel in both parts of the cont-
pounds (as they do in isolation: f»[u:], /??[a:]ðttr, Á:[o:]/?a) while we actually get them
only in the first parts, as far as I know. Again, something more needs to be said about
the interaction of (secondary) stress and length. Similarly, saltkjöt ‘salted meat’ should
have a long vowel in the second constituent (kjöt ‘meat’) but it does not, I think.
In the compounds smáatriöi ‘trifle’ and smáatvik ‘insignificant cvent, trifle’ the a-vowel
of the socond constituent should be long, because it belongs to the first syllable of
the second part of the compound and the following cluster (i. e. /tr/ and /t\f) is tradi-
tionally a lengthening one (or, in Árnason’s terminology, because $ appears between
the two consonants); yet it does not seem that these vowels are long. Exantples of
this sort could be multiplied.
I suggest that two regularities are involved here: the final vowel of a polysyllabic
word (or suffix) is destressed’ and, in compounds, the second of two adjacent stresses
is deleted.
The first rule accounts for the final short vowels in almanak, almanakanna and dttg-
legur and can hopefully be of general applicability in Icelandic word phonology. The
stress deletion rule provides a straightforward account of why the second constituent
of compounds has a short first vowel when the first constituent is monosyllabic: stress
deletion applies before the length rule. In this way a number of apparent irregularities
can be accounted for: the/a/ vowel of maður ‘man’ is short, as noted above, in búmaður
but long in gæfumaður ‘lucky man’. The reason is, of course, that in the latter com-
pound there are no two consecutive stressed vowels (the -fu- syllable separates them),
hence the stress deletion rule is inapplicable and the length rule applies to both mem-
bers of the compound/
To account for the short final stem vowel in alman[a]kanna, we need to take up
the question of the type of boundaries associated with affixes. I suggest, although the
suggestion is admittedly speculative, that some affixes carry with them the word
boundary #. The following argument may be given to support this claim.
As noted by Oresnik & Pétursson (1977:162), the gen. sg. -s ending (frequently)
fails to induce shortening of the preceding vocalic nucleus (or, alternatively, fails to
mhibit lengthening). Thus skips, leiks, báts, — being genitives of skip ‘ship’, leikur ‘play’,
and bátur ‘boat’ — usually have long root vowels and diphthongs. One way of describing
these facts is to separate the desinence from the stem by #. This approach receives
some support from the fact that when the stem-final consonant of skip- undergoes a
3 Alternatively, the stress is weakened to tertiary and the length rule is sensitive
to primary or secondary stress only.
I am thoroughly puzzled by Árnason’s claim that þjóðvísa has a long vowel in
the second member of the compound. In Böðvarsson’s transcription, numerous words
like Þjóðleikhús ‘national theatre’, þjóðnýta ‘nationalise', þjóðrœkinn ‘patriotic’, þjóð-
vtnur ‘patriot’ all have a short vowel in the syllable following þjóð-. If this is the case,
then Árnason’s transcription of þjóðvísa is presumably erroneus and should read [þjouð-
v'sa]. To account for the short diphthong in þjóð- we would need an additional as-
sumption that boundary weakening reduces # to + and the length rule is blocked by
the following consonantal cluster (Árnason achieves this by resyllabification (p. 50)).