Íslenskt mál og almenn málfræði - 2014, Side 120

Hence it would be natural for the speaker acquiring Icelandic to “assume” that there is
some special reason why these words appear to begin with voiceless sonorants. Analyzing
them as containing an initial /h/ would then be the most natural option, especially since
an [h] sometimes appears on the surface according to acoustic analyses. If the Icelandic /j/
is a semivowel rather than a fricative, as has sometimes been suggested (e.g. by Gunnar
Ólafur Hansson 2013:201n), then it is also a sonorant and an apparent voiceless /j/ in ini -
tial position would be equally unexpected as in the other instances. An /hj/-analysis would
then also be a solution to that puzzle.
Finally, it is admitted that alliteration of hv-words with h-words (in some instances, cf.
below) is a more difficult problem. The former are standardly said to begin with [x], [xw]
or [xv]. Accounting for the alliteration here in the same fashion as in the case of other h-
words, one would have to assume an underlying /h/ (plus some sort of a labial element to
account for the [xw] and [xv]-variants). An /hv/-analysis would be a possibility, but then
one might have expected an [f] rather than [x] on the surface. But whatever the correct ana-
lysis of the so-called hv-pronunciation may be, it is clear that those who categorically
pronounce hv-words with an initial [kh] (i.e. have the so-called kv-pronunciation) have re -
analyzed the hv-words as beginning with a /k/. Hence they only hear those words allite-
rate with other words with initial /k/ and not with other words beginning with an h-.
Thus it cannot be the case, as assumed by Haukur (e.g. 2013a:67), that certain speech
sounds form an equivalence class in metrics “because” speakers “hear” them alliterate
together even when they are phonetically and phonologically different in their language.
Speakers do not. It must be the othe way around, namely that speakers hear certain speech
sounds alliterate together because they form an equivalence class from a phonological
point of view. This means, then, that purely poetic and non-phonological traditions cannot
play as big a role in the living Icelandic alliteration as Haukur assumes, although they may
very well do so in societies where poetic traditions are dead and where aspiring verse-makers
have to learn these traditions from books or by other means of scientific study.
Höskuldur Þráinsson
Íslensku- og menningardeild
Háskóla Íslands
IS-101 Reykjavík, ÍSLAND
hoski@hi.is
Höskuldur Þráinsson120