Rit (Vísindafélag Íslendinga) - 01.06.1970, Blaðsíða 481

479
be preferable since it postulates no modification of /D, G/, i.e.,
/ð, y/ in the most general environments (the fricative mani-
festations being found not only in intervocalic and final
positions but also, for example, before /s/ and even before
/t/ in forms like [kb:’yt]). However, the latter solution (i.e.,
phonemic stops) has two advantages. First, it assigns the same
features to /b, d, g/ in strong position and /B, D, G/ in weak
position, whereas these are phonemically different if the
latter are fricatives in the underlying matrix. (Note, however,
that the total number of distinctive features is not affected
since /f, v/ must be distinguished as fricatives from /p, b/
anyway.) Secondly, the latter solution makes the alternations
between [d, g] in strong position and [ð, y] in weak position
found in some foreign words (filologi vs. filolog, etc.; see § 2
above) fit into the general framework without requiring any
additional rule. According to the former solution (i.e.,
/D, G/ = /ð, y/) an additional weakening rule must be set
up in order to take care of these alternations. A rule common
to both native and foreign words can be obtained only if
[b, d, g] in strong positions are supposed to be generated
from underlying fricatives. One may then set up a general
closing rule:
'+obstruent' r n
— CONTINUANT / SP
+ VOICED LABIAL (— voiced)
This rule is, however, intuitively wrong, as it turns the whole
thing upside down.
From a general point of view it is apparently profitable to
generate [ð, y] from /d, g/ rather than generating [d, g] from
/ ð, y/. However, this may be true only as long as both strong
and weak positions are taken into consideration. If only weak
positions are considered, i.e, the positions in which the
morphological and derivational processes of native words
occur, it may be simpler to have the rules work the opposite
way, from /ð, y/ to [d, g]. It may not be quite meaningless to