Rit (Vísindafélag Íslendinga) - 01.06.1970, Page 494

492
The variation between (a)rj and (a)n in some French loan-
words such as chans, elegant, between ay and en(t) as in kompli-
máng-.komplimentéra, encyklopedi suggests other rules. The conditi-
ons are very complicated since the assimilation of these words
varies, and the data need detailed investigation. One might sug-
gest underlying forms with (a)ng and later optional rules which
apply only to certain marked words, giving them (e)n, or vice
versa. In this connection the important thing is that these
marginal cases presumably can be handled by rules and do
not motivate setting up a morphophoneme y.
(2) The Sound g. The derivation of g and / is naturally
associated, and once g is derived / causes few problems. The
rules suggested below reflect the assumed diachronic develop-
ment of g and / (for differences, see below).
Some morphophonemic relations between k and g can be
found, e.g. kamp : kámpa, kall: köld: kyla, kát: káttja. But using
one underlying stem, e.g., in kall: köld : kyla must certainly
imply the reflecting of both ablaut and umlaut in the syn-
chronic description of Swedish, which might seem unnatural.
The relation between k and g is better supported by the varia-
tion in pronunciation, e.g., of kilo, kánguru, káx.
Considering these facts one might suggest the rule
k ~yf / —VPAL
(cf. Weinstock 1970). But it seems reasonable to take into
account another occurrence of a p-like sound, viz., in pjás,
fjásk, moatje, etc. We will therefore suggest that the basic
source for g is j devoiced after voiceless consonants, some of
which are generally deleted. Notice that the only remaining
initial clusters containingy (g) are pj,fj, bj, mj, nj (i.e., labials
or nasals preceding j). The following rules would then be
needed:
This rule would insert j between k, g (palatal stops) and