At 2:39:52 PM on Saturday, December 3, 2011, Hrafn wrote:

> My teacher argue that the vowel triangle for vowels in
> final syllable was /e/, /o/, /a/ before 1200. He say the
> oldest text are pretty consequent about using e and o,
> where later icelandic use i, u.

This is true: they are. (By the way, Dan. <konsekvent> =
Eng. <consistent>, not <consequent>.) However, it's not
clear whether this is an indication of pronunciation or just
a scribal convention. Here, in my translation from the
German, is what Gutenbrunner says in his Historische Laut-
und Formenlehre des Altisländischen:

Since the original sound was /i, u/, cf. spellings like
<foþur> (208), <fotum> (215), <oþali sinu> (219) on
inscriptions with o-rune, and the Modern Icelandic
dialects show no trace of a time with /e, o/, there is a
question whether writing <e>, <o> was a purely
orthographic rule of the written language, or whether /e,
o/ were pronounced in the 12th century. It is in any case
striking that the pronunciation /e, o/ disappeared so
quickly and so completely, and this speaks in favor of a
feature of the written language. On the other hand, the
'hiatus' in the u[m]l[aut] tendency §36f. could be
considered an indication that already at some earlier time
in the development of the Nordic languages unstressed /i/
had for a relatively short time so open a pronunciation
that it did not cause u[m]l[aut].

I've managed to track down the runic citations in Rundata:

(208) is Br Sh3:

§A ...þi---- (+) -ftir + foþur (:) sin (:) þurbio-...
§B f

§A ... [e]ptir fǫður sinn Þorbjǫ[rn].
§B ...

§A ... in memory of his father Þorbjǫrn.
§B ...

(215) is N 540:

§A furu- trikia frislats a
§B uit auk uiks fotum uir skiftum

§A Fóru[m] drengja Fríslands á
§B vit, ok vígs fǫtum vér skiptum.

§A We travelled to meet the valiant men of Frisia
§B (and) we divided the spoils of the fight.

(219) is N 210:

aRintr × karþi × kirkiu × þesa × kosunr × olafs × hins ×
hala × a oþali × sinu

Eyvindr gerði kirkju þessa, goðsonr Ólafs hins
Hála/Halla/Helga, á óðali sínu.

Eyvindr, godson of Ólafr the Slippery/Crooked/Holy made
this church on his allodial land.

> He say, that in 600 - 700 while the i-mutation still was
> productive, old /e/, and old /i/ was still separated. But
> after the i-mutation became unproductive, they merged into
> /e/, and old /u/ and /o/ merged into /o/. And then first
> later, they was raisen to /u/ and /i/.

> That do explain why we have words with /i/ without
> i-mutation: hundi, landi, faðir

It's not needed in order to explain any of these. Both
<hundr> and <land> are a-stem nouns, in which the dative had
the ending *-ai in Proto-Germanic. This became *-ē in
Proto-Scandinavian, which was then shortened to *-e and
never triggered i-umlaut.

<Faðir> has a different story. Pre-Germanic had *faþēr,
still with stress on the second syllable. By Verner's Law
this became Proto-Germanic *fáðēr. The *ē seems to have
been very open, perhaps better represented as *ǣ, since it
subsequently became *ā in Proto-Northwest-Gmc. (Cf. runic
<swestar> 'sister' in the inscription N KJ76 U, dated to ca.
400 CE.) In Proto-Scand. it was shortened (e.g., *faþar
'father') and then raised, first to /e/ and later to /i/,
apparently not early enough to trigger i-umlaut.

It's also worth bearing in mind that umlaut was sometimes
reversed by analogy with non-umlauted forms in the same
paradigm.

I'm not saying that he's necessarily wrong, but the matter
isn't really very clearcut, and the evidence admits more
than one interpretation.

Brian