It's an interesting theory and I wouldn't mind discusisng it if it you're also interested.

Something I think that is important to address is that "semantic shift" here means "the English translation that Bethesda gave the word," not anything internal to the language. Nivahriin means nivahriin, nikriin means nikriin. "Cowardly" and "coward" are approximations that we have to dance around.

Most of the relationships between words we know are either direct compound words or words with straightforward prefixes or suffixes. You're right that it could be a case of confirmation bias - it may seem that the relationships are straightforward because the only ones we know for certain are straightforward.

I would be happy to further discuss the etymology of nivahriin. Unfortunately I'm not sure if anything more can be said about it, given the lack of information.

Would you like to elaborate on your point regarding the meaning of semantic shift? I feel that at some level we have to take the translations given at more-or-less face value, or else try to explain why we don't. Otherwise the whole "game" of etymological analysis becomes impossible.

Here are some roots that appear to exhibit vowel variation:

e/u:

DRE 'do'

Dreh (verb) Do / Does; Drey (verb) Did;
Drun (verb) Bring/Brought, also to cause, usher, or make happen;

ei/e

KEIN 'war'

Kein (noun) War;
>> Vukein (noun) Combat;
Kendov (noun) Warrior

o/u:

HO 'hear'

Hon (verb) Hear;
Huzrah (verb) Hearken

JO 'human, mortal'

Joor (noun) Mortal;
Jul (noun) Man / Mankind / Humans / Humanity;
?  Jun (noun) King, Light; Jud (noun) Queen

o/i

DOV 'dragon'

Dov (noun) Dragonkind, the race of dragons;
Dovah (noun) Dragon, (pronoun) Me; {various compounds on dovah~};
Diiv (noun) Wyrm

It's a complicated topic. I'm of the opinion that we shouldn't take Bethesda's translations at face value, and that the words themselves hold the most accurate meaning.

For example, the word zeymahzin is given to mean "companion". However, if we look at the word itself, we can see that it's a compound of zeymah "brother" and zin "honor," roughly "honored brother." Given the context of the word "Companion" in the rest of Nordic mythology, it becomes clear that zeymahzin has a much more refined definition than the English "companion." A friend could be a "companion" or a dog could be a "companion." A friend could be a zeymahzin, but a dog, probably not.

Similarly, when we look at nivahriin "cowardly," we can see that it's ni "not" and vahriin "sworn" (suppose for this discussion that this is true). A friend who shies away from a drinking challenge could be "cowardly". A dog that deserts you when you encounter a bear could be "cowardly." The same friend is probably not nivahriin, since he or she is not sworn by oath to prove their drinking ability, but the dog is nivahriin since it's supposed to protect you from the bear.

For obvious reasons this exercise doesn't work as well with words that have no discernible roots, but we can still use context to determine what connotations a word has.

Nok means "to lie" and is used in a number of Word Walls to declare which dead person is buried there; Het nok kopraan do Helga "here lies the body of Helga." When we look at related words we might stumble across dinok "death." We may not know the exact relationship between nok and dinok, or if there is one at all, but based on its usage, we can surmise that nok probably refers exclusively to the dead or burial. A dead king can nok, a sleeping person might metaphorically nok, but the sword on the table probably doesn't nok.

It's an exercise in treating the language as its own entity, independent of the English translations that are given. It makes etymology in some ways more difficult, but also helps open our eyes to relationships that we might not have noticed if we were only paying attention to the English translations.

It seems to me what you are describing is essentially the concept of semantic spans. It is no secret that words have different semantic spans in different languages. To take a simple example, the Swedish word löv can usually only mean 'leaf of a plant', whereas the English cognate leaf can mean 'leaf of a plant or a book'. Swedish blad meanwhile can mean 'leaf of a plant or a book, or blade of a knife', while English blade usually means 'blade of a knife', sometimes 'leaf of grass'. Different languages, different semantic spans.

I can well imagine that nivahriin has a different semantic span than English 'cowardly'. I'm afraid this is not enough to convince me that the word must derive from 'not-sworn', however.

As for nok, I would not be to sure that it cannot be used about a sword on a table. Simple concepts like 'lie' tend to work themselves into all sorts of expressions, and the fact that nok is only attested in association with death can probably be explained by the limited extent of our canon. I would accept this idea more readily if there was another word for 'lie' that was attested with no connection to death. Are there any other pairs of verbs with the same translation where each is used in a different context?

There are a few nouns that would appear to be identical without further investigation. For example, shul and krein are both given the translation of "sun." Shul is a Word of Power while krein is not, and shul may relate to sul "day," giving it stronger relations to daylight, heat, or fire (which makes sense given that it's in the Fire Breath Shout).

Another similar case is sunvaar and sivaas, both of which are attested to mean "beast." Sunvaar is used in reference to a vicious werewolf, so sunvaar is more specifically "a ferocious beast," while sivaas is used in reference to Kyne and friendship with animals, so it is probably closer to "beast of the field/creature/animal" in a more natural sense.

Jul and mun is another good example. Jul means "man" as in "mankind/the races of man," while mun refers to a singular "man/person." It's the same distinction between dov and dovah. This one is commonly confused among beginners.

@paarthurnax, thank you! Good examples all of them.

~Zul