Íslenskt mál og almenn málfræði - 01.01.1984, Síða 148

146 Kristján Arnason
nonexistence of forms like hahh [hah:], compared to hatt [hah$], vaðð
[vað:], compared to vaðs [vaðs] and [haj-:] compared to hart [haj^]
etc. creates a gap in the pattern. One way of dealing with these gaps
is to consider them as accidental in the sense that tradition does not
supply lexicogrammatical material for the forms in question to repre-
sent. In fact, a limited number of innovative forms are to be found
which might be seen as tending to fill these gaps. There are exclama-
tions like hah [hah:] and puh [phYh:] which may be considered to
enter into the pattern as having an empty coda filled by a copy of
the nonvocalic part of the nucleus (or else with underlying [h] in the
coda). Similarly, forms like meðða [með:a] ‘tig’ and Sivva [slv:a] ‘a
woman’s nickname for Sigurveig', may be seen as filling the gap
provided by the possibility of a voiced nonvocalic nucleus followed
by a copied or underlyingly identical coda. In the light of these con-
siderations, the account in terms of an underlyingly empty coda filled
by a rule copying the consonantism may seem to become attractive.
There is some evidence that this ‘rule’ is ‘productive’ in the phonology
of Modern Icelandic in that acronyms like SÚM (Samband ungra
myndlistarmanna ‘The Young Artists’ Union’) and SÚN (Samvinnufé-
lag útgerðarmanna á Neskaupstað ‘Neskaupstaður Shipowners’ Coop-
erative’) where ‘lexically’ there is only one consonant, may be pro-
nounced [sum:] and [sun:], which can be accounted for by assuming
that the phonological transformation supplies the additional conson-
antal material lacking in the lexical representation. (I will return
shortly to the question of the ‘productivity’ of rules.)
Also relevant to the question of the relation between the coda and
the nucleus is the fact that there are certain combinations of nuclei
and codas that are predicted by the model but do not occur, e.g.:
Vþn, Vnl, Vxl, Vþl, Vrl, Vxl, Vxn, Vhs (i.e. preaspiration before
/s/), Vhl, Vhn, Vhr. These are mostly nuclei with checked voicing
followed by sonorants. I would not want to make a bet at the present
as to whether these gaps are accidental or whether they are
‘phonologically principled’. Other gaps that are found are easily ac-
counted for, such as the one created by homorganic nasal assimilation,
excluding combinations like: Vnþ, Vq$ etc.4 Further, [s] in the coda
4 It is here assumed that forms like þyngd ‘weight’ [þirjd] are derived from underlying
/þingd/ by a rule that deletes the velar stop in front of a consonant after velarization