In an effort to show how an historical linguist does their job, I will demonstrate the evidence behind Proto-Germanic phonology from the standpoint of Old Icelandic and its cognates. These lists are out there in the world. But in addition, I will list exceptions to these rules as well. I will not use established sound laws to complete this. That would be cheating. Instead, I will only use the data at hand to draw conclusions. I shall start with the initial <a> in Old Icelandic.
Old Icelandic a- / au-initials
In all instances, Old Icelandic a-initial corresponds to Gothic a-initial, even in the diphthong au-initial. The harmony and consistency is compelling.
|agis, agei||agi||"terror", "uproar"|
|at||at||"at", "by", "against"|
|auk||auk||Goth. "for", "also", "but"; OIc. "besides"|
|ahs||ax||"ear of corn"|
There are no exceptions to these cognates, demonstrating the remarkable congruity between Old Icelandic and Gothic as a- and au-initials point to Proto-Germanic *a- and *au-initials as well.
With West Germanic languages, the picture becomes far murkier. Because there are many more cognates between Icelandic and West Germanic languages, I will not list every example.
Old High German and Old Saxon
|af||aba, ab||af||"off", "from"|
|angr||angar||OIc. "bay", "harbor"; OHG "meadow"|
|argr||arg, arag||arch (MLG)||"effeminate"; "lewd"; "wicked"|
|ari||aro, arn||are, ar (MLG)||""|
|ax||ahar||"ear of corn"|
While most cases of OIc. a-initial corresponds to an a-initial in OHG or OS, we lose the universal correspondences we saw between OIc. and Gothic. In OHG, there are exceptions where a is raised to e. I shall include Gothic examples in these exceptions as well.
|aðili, -ja, -jar||ethili||"chief defendant", "prosecutor"|
|agi||agis; agei||egī||"terror", "uproar"|
|alin||aleina||elena, elna||"half yard"|
|alri||erila, elira||"elder tree"|
|annarr||anþar||ōthar, āthar, andar||"other", "second"|
The obvious answer to many of these examples is the influence of e and i in the second syllable causes the a-initial in OHG and OS to mutate to e-initial. In examples for "chief defendant," "terror," "half yard," "inheritance" and "elder tree," the Proto-Germanic sound in the second syllable must have been either *ī or *ei. If it were *ī, OHG elena/elna "half yard" points to a Proto-Germanic *alīn- that became *elina in pre-OHG and finally influence from the stressed e-initial influenced the second syllable to elena.
When a Proto-Germanic first syllable had *aCC- (C standing for a consonant of any value) and OS or OHG reduced it to *aC-, it compensated the reduction to a single consonant by lengthening the a-vocalism to a long ā. This is proven in OIc. annarr "other" ~ Gothic anþar ~ OS ōthar, āthar, andar, where when the consonants in Old Saxon were not reduced (cf. andar) the vowel was not lengthened, and is reinforced by OIc. aptann "evening" ~ OS āƀand and OIc. arfr "inheritance" ~ OHG āband, ābund (with an -and suffix in OHG that is not present in OIc. or OS). I believe the Old Saxon long-o counterexample ōthar is due to simultaneous influence from (1) the movement from back to front because of the subsequent dental th and (2) analogy with other ō-initial words, arising when au-initials rose and merged to ō (see below). The confluence of both the fronting movement in the mouth and the sound change of au- to ō- changed the word in freak occurrences.
In OIc. au-initials, which correspond to Gothic au-initial, OS merged and lengthened the diphthong to a long monophthong ō. In OHG, the au- rose to ou- but only merged and lengthend to ō- if followed by a dental. If the following consonant was in the back of the mouth, ou- was preserved. This novel split has to due with the movement in the mouth. s an independent word, au was preserved in OHG (seen in German au Weh).
|auð-||ōdi||ōthi||augmentative prefix, "easy"|
|auðn, auðna, auðr||ōdi||"wilderness", "desert", "void"|
|aufi, auvi||au Weh (Germ.)||interj. "woe!"|
|auk||ouh||ōk||adv., conj. "besides"|
|ausa||ōsen, œsen (MHG)||"east"|
Now *a- was nasalized before *n. Proto-Germanic cluster *anþ- was voiced to *anð- in West Germanic. Old Saxon dropped *n and lengthened *a- to *ā- (we saw this happen in āƀand), but because *ā- in this instance was nasalized, Old Saxon raised it to *ō-. The process was incomplete in Old Saxon; evidently other dialects preserved older forms for the word. This route of n-dropping and raising a nasal is also seen in Old English and Old Frisian (below). None of this happened in Old High German.
|annarr||andar||ōthar, āthar, andar||"other", "second"|
Cognates are scant and most are found in Middle or Modern Dutch, when later sound changes had further distorted the language. To keep things neat, we will only consider Old Dutch examples. We can see that sound changes to the a-initial in Old Dutch mirror sound changes in OS. The default status of a- is unchanged.
Subsequent consonants or vowels can condition the status of a- and change it to a long-a or a long-o, as we saw in OS. I will include their lemmata when applicable. Long-a arises if the following consonant is dropped. Long-o if *au- merged.
|auk||ōk||ōk||adv., conj. "besides"|
|auka||geōcodion||ōkian||OIc. & OS "augment;" ODu. "cause to increase"|
There is one short-o example as well, a puzzling exception to everything we have seen thus far:
|ausa||osen||ōsen, œsen (MHG)||"sprinkle", "pour"|
I included the cognate form from OHG, where we see an unusual development as well. The expected form in ODu. is **ōsen. I have no explanation for this.
Old English and Old Frisian
Old English and Old Frisian are part of the Anglo-Frisian sub-branch of West Germanic. We can see from the onset why that is. We see a- in Icelandic correspond with a- in English or Frisian, but we do not see the consistent harmony in Gothic or even OHG.
Spelling in Old English was chaotic. It's easier to speak of initial phonemes than what is written on the page. Nevertheless, we cannot change the "data" (the letters the writers chose). The first set of correspondences are between OIc. a- and a short "a" (either /ɑ/ or /æ/) in OE and e- in OF. We see that <a> could be written as <æ> in "ash tree."
|alda||aldaht, aldot||"heavy wave"|
|alri||alor, aler||"elder tree"|
|askr||asce, axe||"ash tree"|
|atall||atol||OIc. "fierce"; OE "terrible"|
|at||æt||et||"at"; "by"; "against"|
Sometimes a-initial corresponds to e- in Old English. These examples mirror a change in the West Germanic languages, where if the vowel in the second syllable is an <i> it will mutate the a-initial into an e-initial (cf. OHG elena, elna and Middle Dutch ellen, elne "half yard"). You may note that OIc. alri corresponded with OE alor, aler "elder tree" without an e-mutation in English (OHG erila, elira and OIc. variant elri show a mutation). The Old English form points to a pre-form without an -i, *alr, while OIc. and OHG point to *alri.
Before a liquid consonant, OIc. a- corresponds with OE ea-/ie- and OF e-. Confusingly, OE <ea> represented /æa/ while <e> represented /e/ (the variation between eal and ǣl confirms this). So what is written as <ea> is a vowel breaking of <æ> /æ/, not <e>; the pre-form of words like eall "all" was probably closer to *æll.
|allr||eall||*el- (cf. elmachtig "almighty")||"all"|
|arta||earte||OIc. "teal"; OE "wagtail"|
OIc. au- corresponds to OE ēa- / īe- and OF ā-. Fairly simple merger and compensatory lengthening that we've seen before, with vowel breaking as well in the English. For "ear of corn" we see a great deal of spelling variation, probably attempting to represent the long consonantal cluster.
|auk||ēac||āk, ēk||adv./conj. "besides"|
|aurr||ēar||"earth", "mud", "clay"|
|aurr||ēar, æhher, eher||ār||"ear of corn"|
Initial apt- in Old Icelandic corresponds to OE ǣf- and æft-, pointing to different proto forms in Germanic. Frisian iō- is unexpected.
Now onto the oddballs. In the following example, evidently the weak a- became a weak back vowel or perhaps a schwa. We saw this happen in OHG az above, where the weak vowel could be further distorted to ez and iz.
In the next example, we see the opposite of what happened with alri ~ alor "elder tree," where the Old English preserved a form without -i that was added secondarily. In the following case, the Old Icelandic form preserves a word without -i, while -i caused the expected mutation of a-initial to e-. The pre-OE consonant cluster *enð- was then reduced, and the vowel underwent compensatory lengthening, which happened in Old Saxon too.
The final example mirrors what we saw in Old Saxon and has the same explanation.
So our study shows the remarkable conservativism of Old Icelandic and Gothic. It also reveals the sound changes, both expected and not in other languages.
|a- (incl. apt-, Type I)||a-||a-||a-||a-||a-/æ-||e-|
|a-, weak vowel||a-||a-/e-/i-||a-||a-||o-/æ-||o-/a-|
|a-, i in 2nd syllable||a-||a-/e-||e-||e-||e-||e-|
|an-, i in 2nd syllable||an-||ē-|
|apt-, Type II||āƀ-||āv-||ǣf-||ēw-/iōw-|
|auC-, back consonant||auC-||ouC-||ōC-||ōC-||ēaC-/īeC-||āC-|
You could probably program these rules and create a word transmogrifier.