Since there appears to have been little done on the subject of etymological analysis, I've decided to make an attempt. I would here like to present my preliminary findings. I apologize beforehand if anything I address has already been discussed extensively. I have actively avoided searching for earlier studies, so as to approach the subject with as few preconceptions as possible. No doubt you will find many errors or contestable points. I hope that some of you will find this amusing, and that we might work together to make corrections and further discoveries.

I wish to stress that everything below this point is speculative! When I use the word "is" below, I most likely mean "seems to be". Nonetheless, I believe that most of my speculations rest on fairly solid, although horribly thin, foundations.

Roots and stems

First, some boring definitions. If you know your linguistics you may skip the next two paragraphs.

Roots in linguistics are abstractions: the smallest common denominators from which all related words can be derived by various methods. An example of a root in Proto-Indo-European (an ancestor of English and many other languages) is DEL² 'recount', which yields a large number of related English words including 'tell', 'tale', 'talk'. (The superscripted 2 means that there is also an unrelated DEL¹ 'long', the origin of words such as 'linger', 'Lent' and 'lunge' by a more complicated derivational process).

Roots are not words, even though they may appear identical to simple words (and indeed, in Dovahzul many simple words are indistinguishable from their roots, as might be expected from a language supposedly close to its own origins). As such, roots do not have translations in themselves, although in the presentation below I usually provide tentative interpretations of the root "meanings" for the sake of convenience.

Roots in Dovahzul have the configuration (C)CV(V|h) or (V)VC|h, where C is a consonant, V is a vowel, h is the consonant H and anything within parentheses is optional. We thus see roots such as BO 'move, transfer, give'; BRI 'beauty'; DII 'go cold'; DWII 'carve'; FAH 'favouring'; OV 'trust'; AH 'wound'; IIZ 'ice'. Some words are derived directly from roots, such as bo 'fly, move'; fah 'for'; ov 'trust'; iiz 'ice'.

Stems are extensions of the roots, derived from them in a few different ways. The most common way is to attach a consonant ending to a root. This ending can be

D; F; FT; G; K; L; LD; LT; M; N; ND; NG; NT; P; Q; R; RK; S; ST; T; TH; V; X; Y; YT; YTH; YV; Z

that is, any of the consonants in the dragon alphabet except B; H; J; W, but including TH; Y and the consonant clusters LD; LT; ND; NG; NT; RK; ST; YT; YTH; YV.

By this derivational process we arrive at stems such as BRI-T 'beautiful' and BRII-N 'beautiful one'; DI(I)-N 'freeze, die' and DI-R 'die'; DWII-N 'steel' and DWII-R 'carve'; AH-R 'hurt' and AH-Z 'painful'. Many words are based directly on stems ("stem-words"), such as brit 'beautiful'; diin 'freeze'; dir 'die'; dwiin 'steel'.

Another, less common, way of derivation seems to be to reverse the root. Possible examples include UN 'un-' vs. NU '-less'; THU 'power' vs. UTH 'command'; GO yielding gogil 'goblin' vs. OG yielding ogiim 'orc'.

A third derivational mechanism we might call consonant and vowel variation.

The consonant variations have to do with the initial consonant S and the final consonant H. For S variant roots occur with F, SH, TH and Z. So SU-L 'day'/SHU-L 'sun'; SU/THU 'air'; SU/FU 'power'; SAA/ZAA 'lose'. Regarding H, this seems to be optional in some roots, as suggested by the pairs SA/SAH 'protect'; SO/SAH 'blood/red'. This might also explain the variation in balaan / bahlaan 'worthy'.

The vowel variations are more numerous:

• AA/E: DAA/DE 'fate'
• A/O: SA/SO 'pale'
• AA/II: LAA/LII 'end'
• AA/I: STAA/STI 'free'
• E/EI: KE/KEI 'commotion'
• E/I: NE/NI 'not'
• E/O: RE/RO 'rise, branch out'
• I/II: BRI/BRII 'beauty'
• II/U: DII/DU 'swallow'
• O/OO: LO/LOO 'have'
• O/U: BRO/BRU 'nord'

Derivational affixes

Further words are constructed by adding derivational affixes to either roots or stems. Dovahzul has a large number of derivational affixes (between 80 and 90), most of which are attested in just a few words, and few of which can be confidently analysed. It seems clear, however, that the choice of affixes is rarely dictated by word classes (at least not as we understand them). There are some affixes with clearer meaning, though:

• Suffixes with ii seem to be preferred for race-words. The suffixes -(i)il and iim are used exclusively for "races": ogiim 'orc'; fahliil 'elf'; gogil 'goblin. The pattern is reinforced by the suffix -iik, which is found on grohiik 'wolf', ufiik 'troll', bruniik (adj) 'savage' and reyliik 'race/heritage'.
• The suffix -maar seems to have to do with the sense 'self, person', being used in the words deinmaar 'keeper' (from dein 'keep'); geinmaar 'oneself'; nimaar 'itself'
• The prefix ah- seems to be used to form nouns from adjectives: ahkrin 'courage' / krin 'courageous'; ahmul 'husband' / mul 'strong'; *ahzah 'end' (reconstructed from unahzaal 'unending') / zah 'finite'. Likewise, ahrol 'hill' possibly derives from an unattested adjective *rol 'rising' (from the root RE/RO 'rise, branch out').

Returning to the stems used in the examples above, the stem BRII-N yields briinah 'sister'; DI(I)-N yields dinok 'death'; DWII-R yields dwiirok 'carve'; AH-R yields ahraan 'wound'; AH-Z yields ahzid 'bitter'.

The following lists all affixes with interpretable meaning, based on comparison with stem-words with closely related meaning where the affixes are missing. This does not necessarily mean that the words with affixes derive from the stem-words: they might derive directly from the stem itself, wherefore the stem-words and the derived words do not always agree completely in sense.

The attested examples are usually very few. Therefore all attestations of these affixes are given as examples. Note: many of these affixes also occur in words where a corresponding stem-word is not attested. In these words the significance of the affix might be different.

• a- verb ("towards") from noun
  lok (noun) sky / alok (verb) arise, rise
• ah- noun from adjective
  krin (adjective) courageous / ahkrin (noun) courage
• mu- adjective from abstract noun
  nax (noun) cruelty / munax (adjective) cruel
• r- 3p personal pronoun from possessive pronoun
  ek (pronoun) her, hers / rek (pronoun) she
  ok (pronoun) his / rok (pronoun) he
• si- perfect participle from present verb
  saan (verb) lose, loss / sizaan (adjective) lost
• vu- specific noun from general noun
  kein (noun) war / vukein (noun) combat

• -aag abstract noun ("the state of being ~") from adjective
  mul (adjective) strong / mulaag (noun) strength
• -ah
  1. specific noun from general noun
     dov (noun) dragonkind / dovah (noun) dragon
  2. noun from noun
     mon (noun) daughter / monah (noun) mother
  3. adjective ("~-like") from noun
     gol (noun) stone / golah (adjective) stubborn
  4. verb ("use a ~") from noun
     horvut (noun) trap / horvutah (verb) trap
• -aal
  1. agent noun from abstract noun
     kah (noun) pride / kaal (noun) champion
  2. verb ("lose ~") from noun
     sos (noun) blood / sosaal (verb) bleed
• -aan
  1. agent noun ("characterised by / performing ~") from noun or verb
     kah (noun) pride / Kaan (noun) Kyne; dukaan (noun) dishonor / dishonored
     kriivah (noun) murder / krivaan (noun) murderer
     tovit (verb) search / tovitaan (noun) searcher, seeker
  2. noun ("significant male") from noun (counterpart of -aas 1)
     kul (noun) son / kulaan (noun) prince
  3. abstract noun (state, "-dom") from adjective
     onik (adjective) wise / onikaan (noun) wisdom
  4. active verb ("again-/turn-") from verb or noun
     prodah (verb) foretell / dahmaan (verb) remember
     grah (noun) battle / graan (verb) rout (flee)
  5. past verb from present verb
     wahl (verb) build, wahlaan (verb) built
  6. perfect participle from verb
     mah (verb) fall / fell / mahlaan (adjective) fallen
  7. "superlative" adjective from adjective
     pruzah (adjective) good, well / pruzaan (adjective) best
• -aar
  1. noun (domain or state) from noun or adjective
     jun (noun) king / junaar (noun) kingdom
     kras (adjective) sick / krasaar (noun) sickness
  2. agent noun from noun
     sahsun (noun) village / sahsunaar (noun) villager
  3. adjective ("-ful") from noun
     suleyk (noun) power / suleykaar (adjective) powerful
• -aas
  1. noun ("significant female") from noun (counterpart of -aan 2)
     kul (noun) son / kulaas (noun) princess
  2. noun from noun
     lah (noun) magicka / laas (noun) life
• -drim noun from noun
  hah (noun) mind / hahdrim (noun) mind
• -ei 2p inflection verb
  los (present copula) is / losei (verb) are you
• -ey pt inflection verb
  dreh (verb) do / drey (verb) did
• -hus abstract noun (state) from concrete noun
  in (noun) master / inhus (noun) mastery
  zaam (noun) slave / zaamhus (noun) slavery, servitude
• -iid agent noun ("one who ~") from verb
  kron (verb) conquer / kroniid (noun) conqueror / victor
• -ik passive participle from verb
  daan (verb) doom / daanik (adjective) doomed
• -in agent noun ("one who ~") from verb
  qahnaar (verb) vanquish / qahnaarin (noun) vanquisher
• -iv abstract noun ("being ~") from adjective
  onik (adjective) wise / onikiv (noun) enlightenment / enlighten
• -kei adjective ("having ~") from noun
  moro (noun) glory / morokei (adjective) glorious
• -l noun ("characterised by ~") from adverb
  ni (adverb) not / nil (noun) void
• -maar
  1. agent noun ("one who ~") from verb
     dein (verb) keep / deinmaar (noun) keeper
  2. reflexive noun ("-self") from noun or pronoun
     gein (noun) one / geinmaar (pronoun) oneself
     nii (pronoun) it / nimaar (pronoun) itself
• -n plural form of pronoun
  nii (pronoun) it / niin (pronoun) them
• -naar verb ("successfully using ~") from noun
  qah (noun) armor / qahnaar (verb) vanquish
• -o passive participle from verb
  naak (verb) eat / naako (adjective) eaten
• -om abstract noun from adjective
  vul (adjective) dark / vulom (noun) darkness
• -on
  1. noun from adjective or preposition
     vul (adjective) dark / vulon (noun) night
     fah (preposition) for / fahdon (noun) friend
  2. verb ("act as a ~") from noun
     diiv (noun) wyrm / diivon (verb) swallow
• -ot adjective from adjective
  kril (adjective) brave / krilot (adjective) valiant
• -os specific noun from general noun
  yol (noun) fire / yolos (noun) flame
• -raak noun ("state of being ~") from adjective
  hevno (adjective) brutal / hevnoraak (noun) brutality
• -raan specific noun from, or product of, general noun
  kip (noun) food / kipraan (noun) meal
• -rah verb from verb
  liiv (verb) wither / liivrah (verb) diminish
• -sun agent noun from verb
  vey (verb) cut / veysun (noun) ship
• -t
  1. past participle from verb
     kren (verb) break / krent (adjective) broken
  2. past copula from copula
     los (copula) is / lost (copula) has
• -ul specific verb from general verb
  bo (verb) fly, flow, move, arrive, come, go / bovul (verb) flee / fled
• -um specific noun from general noun
  su (noun) air / su'um (noun) breath, inner spirit
• -un abstract noun ("state of being ~") from adjective
  mid (adjective) loyal / midun (noun) loyalty
• -us adjective from verb or noun
  motmah (verb) slip / motmahus (adjective) slippery
  od (noun) snow / odus (adjective) snowy
• -vut verb ("use ~") from noun
  haal (noun) hand / haalvut (verb) touch
• -z specific noun from general noun
  gol (noun) stone, earth / golz (noun) stone

Phonotactic rules

In the construction of the derived words in the canon, either by compounds or by the use of derivational affixes, there are few phonotactic rules that seem to come into play. Phonotactic rules make sure that the newly formed words agree with the sound-laws, or phonotax, of Dovahzul, and prevent the emergence of unallowed sounds or sound sequences.

One phonotactic rule seems to be that h must not occur between certain vowels, so when a suffix beginning in a or e is attached to a root ending in h, the h vanishes. For example, the verb lahney 'live' appears to be related to lah 'magicka, mana' and laas 'life'. It seems likely that the root is LAH, and that laas has been derived from it by the well-attested suffix aas. Further support for this sound-rule is given by the seemingly related pairs ahmik 'service' / aar 'servant', gahrot 'steal' / gaar 'unleash'; grah 'battle' / graan 'rout'; kah 'pride' / kaal 'champion'; kriivah 'murder' / krivaan 'murderer', and others. That ehe becomes e is suggested by dreh 'do' / drey 'did'; it appears likely that the suffix is -ey. (Note that the e is short: ee does not seem to occur in the canon.)

There are a number of other sound sequences which Dovahzul does not permit. Many of these are broken by the use of connecting consonants.

The connecting consonant d is used

• Between vowels in compounds: fodiiz 'hoarfrost'
• Between h or a vowel and the suffix -on: fahdon 'friend'; sadon 'grey'
• Between a vowel and the suffix -iis: bodiis 'borrow' (also tahrodiis 'treacherous'?)

The connecting consonant l is used

• between y and a suffix beginning with a vowel (suggested by reyliik 'race').
• between h and the suffix -aan in adjectives: bahlaan 'worthy'; mahlaan 'fallen'. (But note mulhaan 'unmoving'. Here lh makes the connecting consonant unnecessary, and probably also prohibited.)
• possibly between h and suffixes beginning in -o: bahlok 'hunger'; nahlot 'silence'; pahlok 'arrogance'; sahlo 'weak'. (These might also derive from stems in -L, but it is notable that none of them have related words based directly on such a stem – nahl 'living' appears to be unrelated to nahlot 'silence'.)

The connecting consonant m is used

• between h and the suffixes -aan or -iik in verbs: dahmaan 'remember'; zahrahmiik 'sacrifice' (contrast grohiik (noun) 'wolf')
• between h and the suffix -in: vodahmin 'unremembered'
• between h and the suffix -ik: ahmik 'service'

The connecting consonant n is used

• Between the prefix si- and a vowel or d: sinon 'instead'; sindugahvon 'unyielding'.

The connecting consonant s is used

• Between n and l in compounds: unslaad 'unending'

The connecting consonant v is used

• Between a vowel and a suffix beginning with u: bovul 'flee'; gravuun 'autumn'

The word sizaan 'lost', related to saan 'lose', suggests that s may be voiced to z in certain contexts. As far as I can tell we have no further examples of this, however (and intervocalic s occurs in other words such as krasaar 'sickness').

The genitive ending -ro (seen only in zindro 'triump's'), if related to do 'of', implies that ndd becomes dissimilated to ndr.

A variant of un- 'un-' is uz-, appearing in uznahgaar 'unbridled'. This suggests that n becomes z when followed by another n.

A tentative list of all roots, stems, derivational endings and derived glosses is available as a spreadsheet here.

~Zul

Excellent writeup, Zulfahdon. I'm pinning this thread for others to see.

This is great to understand when considering new words, and huge swaths of the non-canon vocabulary would benefit from this analysis.

i find this to be a great help for me to understand the basics of non-canon word

Truly amazing work, Zulfahdon! Balaan morah. Your discovery of the "root reversal" method blows my mind, but it absolutely is just so. In fact you just made it way more difficult to invent plausible words. I shudder to imagine the overhaul ahead.

Nox, pah! I'm glad you find this useful. Also, I'm sorry for any trouble I've caused for the word-makers ... ;)

I have made two small addendums, adding one bullet-point below "The connecting consonant l" and a paragraph at the end about uz-.

~Zul

I believe rah "god" and raan "animal" may be related given that the ancient Nords once revered animals, but in a way those could be an extension of your re/ro root (must closely tied to rel "dominate").

A note about nahl "living," it comes from nahlaas "alive," which comes from laas "life." It's a perplexing one.

rah/raan: interesting! I have myself associated rah with rahg 'rage' and the root seen in revak 'sacred' (which in turn seems related to the RE/RO root). Raan I have connected to rein 'roar' and also rahg. But it seems entirely plausible that all of these are connected through the RE/RO root (where indeed I include rel 'dominate').

nahl: again, interesting (and somewhat perplexing). I have connected nahl with nah 'fury' through a root meaning something like "movement" – which still can explain the first element – but a derivation from laas of course makes much more sense. (Do we know this connection from an actual source, BTW?)

Are there other examples of derivations arrived at by dropping an ending, as nahlaas > nahl?

No actual source, but I think it's the most straightforward conclusion given laas and nahlaas.

There is one other example, wahl "to make/build/create," and wahlaan "creation."

paarthurnax

No actual source, but I think it's the most straightforward conclusion given laas and nahlaas.

There is one other example, wahl "to make/build/create," and wahlaan "creation."

Are you saying that wahlaan is from laan 'want, request'? Such a connection seems less straightforward than the one between laas and nahlaas. Why not wahl > wahlaan? The ending -aan is after all well known from other words, such as mahlaan 'fallen' < mah 'fall'.

Zulfahdon

Are you saying that wahlaan is from laan 'want, request'? Such a connection seems less straightforward than the one between laas and nahlaas. Why not wahl > wahlaan? The ending -aan is after all well known from other words, such as mahlaan 'fallen' < mah 'fall'.

Nope, I very much agree with you that wahl comes from wahlaan  and has nothing to do with laan. From a meta perspective, I think wahl and nahl are both situations where Bethesda took a word and removed the ending, and in the case of nahl neglected to remember that nahlaas comes from laas. How we reconcile this in-language I'm unsure.

In English, it'd be as if you had the words "alive" and "life," and then made a word "al".

Oh, all right then. :)

I just added to the top post a list of all derivational affixes in the corpus that are easily analysed. A comparison with Foduiiz' List of Affixes might be enlightening, though to avoid bias I have not undertaken it myself.

Zulfahdon, what do you think about the possiblity that the si- in sizaan "lost" could be the same or related to the sin- in sindugahvon "unyielding"?

We know that gahvon is "to yield" and that du- is negative as seen in dukaan "dishonor," so we can state with a fair certainty that sindugahvon is sin- + du- + gahvon. The question then is what sin- indicates, and if it possibly relates to sizaan (possibly si- + saan).

I like that idea, it makes a lot of sense. Si(n)- might then be a participe or adjective prefix. It is interesting that the consonant following i appears to be mutated in both words; perhaps the prefix is actually sin-, turning *sin-saan > sizaan? (But then there is ensosin 'bewitch' where such a mutation does not occur.)

Another word with the same prefix could be sinon 'instead' (technically an adverb). The latter element might then be related to nuz 'but', hinting at a root NU/ON 'contrary/different' (apparently the same as NU 'not'). And that would in turn explain the final element of dilon 'undead'!