Íslenskt mál og almenn málfræði - 01.01.1979, Page 250

230
Janez Oresnik
Problematic in this connection is jarðneskur [-rðn-] ‘earthly’. (My
attention has been drawn to this example by Eyvindur Eiríksson viva
voce.) At least one word with etymological /rð + n/, harðna ‘harden’,
is now pronounced with [-r$n-]. Both jarðneskur and harðna are old
words,see Larsson 1891 s.vv.Cf. also ha[rQn\eskja ‘severity’ vs. v/[h$n]-
eskja ‘knowledge’. I shall leave the question of jarðneskur open until
more examples of the relevant kind are adduced, and I conclude for the
time being, on the basis of (l)-(2) above, that -nesk- is a derivational
suffix, not a compound-word constituent.
As with rövl(a) above, one could assume that slafnesk- is simply an
exception to the v-to-fo rule. The argument against this runs parallel to
that presented above for rövl(a). Notice that the oldest example of
slafnesk- in the OHÍ is from 1838 (in Fjölnir IV 2, 26; Gunnlaugur
Ingólfsson per litteras). Thus the lexeme has been in the language for
more than 140 years, and there is no indication that there has ever been
any tendency to replace its [-vn-] with [-^n-].
However, this undesirable treatment cannot be remedied with the
help of Kiparsky’s universal in this case, for the lexeme is a derived
form in Kiparsky’s terminology. It contains the root slaf- (cf. Slaf-i
‘Slav’, a loanword), followed by the derivational suffix -nesk; the under-
lying representation is consequently /slav + nesg/, with a crucial mor-
pheme boundary between the /v/ and the /n/. In order that the [vn]
can be explained in a principled way, I suggest that the notion of
DERIVED FORM/INPUT be modified so that segment strings cru-
cially containing a derivational morpheme boundary are not considered
derived if present in the underlying representations. This revision has
been suggested to me by the circumstance that Kiparsky adduces no
examples of derived forms/inputs crucially involving a derivational
morpheme boundary in underlying representations. Given this revision
of the notion DERIVED FORM/INPUT, the lexeme slafnesk- is
straightforward: it is not a derived form with respect to the v-to-b rule,
and consequently the rule does not apply to it. My revision of Kipar-
sky’s universal predicts that new derivations comparable to slafneskur
will keep their [vn] unchanged unless something artificial intervenes.
4. There remains the question of examples such as gotneskur, pro-
nounced with ‘preaspiration’ which contradicts my claim that no neutra-