The etymology of Mari *jŭmǝ ‘sky; god’

In this post I will present a new etymology for the Mari word for ‘sky’ and ‘god’, which can be reconstructed as PMari *jŭmǝ on the basis of MariE jumo, MariNw jŏmŏ and MariW jǝ̑mǝ̑. In present-day usage the meaning ‘god’ is predominant, but also the meaning ‘sky’ is widely attested in Mari dialects. This may be a surprising word to discuss etymologically, as it has a well-established and generally accepted Uralic etymology already.

I’ll first discuss the previously held etymology connecting Mari *jŭmǝ with Finnish jumala ‘god’, and then present an argument supporting the overlooked alternative possibility that Mari *jŭmǝ is a cognate of Finnish ilma ‘air’ (< PU *jilma ‘sky’). In addition, I will also briefly discuss the etymologies of Mari *jŭt ‘night’, *kŭrǝkt-e- ‘to lie, to babble’, and *ŭžed-a- ‘to hate’, as they provide previously overlooked parallels for the vowel development I assume for Mari *jŭmǝ.

The established etymology: Mari *jŭmǝ ~ Finnic *jumala ‘god’

All references (including UEW and SSA) agree that Mari *jŭmǝ ‘sky; god’ goes back to (post-)PU *juma and is related to the Finnic word for ‘god’, Fin jumala, Est jumal, etc. (< PFi *jumala). According to SSA the comparison of the Finnic and Mari words was first suggested in the 18^th century already. In addition, possible Mordvin cognates have also been proposed. First, it has been suggested that MdE jondol, M jondǝl ‘lightning’ (< PMd *jondǝl) is an obscured compound word reflecting Pre-PMd *juma-tuli ‘god-fire’ (cf. PU *tuli ‘fire’ > MdE M tol). Second, it has been proposed that the obsolete Mordvin deity name ‹Jumishipas› could also contain the reflex of the word *juma. The appurtenance of both Mordvin words has been regarded uncertain.

The generally accepted etymology of PMari *jŭmǝ is in itself plausible, although not perfect: there are potential weak points. The etymology is based on a comparison of two forms only, PMari *jŭmǝ ‘sky; god’ and PFi *jumala ‘god’. The etymologies of the allegedly related Mordvin words are speculative, and they do not actually provide any independent evidence for the former existence of the noun *juma ‘god’. On the contrary, only if such a noun can be securely reconstructed on the basis of other evidence, it makes sense to hypothesize that PMd *jondǝl ‘lightning’ could be an obscured compound word containing a reflex of this word. Nothing in the structure of the Mordvin form itself, however, suggests that it was originally a compound noun. As regards the deity name ‹Jumishipas›, it is an obscure hapax legomenon found in Strahlenberg (1730), and as such its etymological interpretation remains entirely speculative; it offers no independent evidence for the former existence of a noun *juma ‘god’.

Thus, the reconstruction of the noun *juma is depends wholly on the assumption that PMari *jŭmǝ ‘sky; god’ and the part *juma- in PFi *jumala ‘god’ reflect the same etymon. Phonologically and semantically the comparison is unproblematic. The only thing in the etymology itself that can be seen as an obstacle is the assumption that a derivational suffix *-lA was added to the word in Finnic. Nothing in the Finnic data, however, suggests that *jumala is a derivative, and there is moreover very little evidence of the existence of a denominal noun suffix of this shape in Uralic. A possible example is found in the word for ‘hare’: cf. SaaN njoammil, MdE numolo, Komi (dial.) ńimal, Hung nyúl (< PU *ńomala) ~ NenT ńawa, Ngan ńomu, SlkTa ńoma (< PU *ńoma). In this case the underived basic form (which is attested in Samoyed) appears to have been augmented with an opaque suffix *-lA that lacks any discernible semantic function, but invoking such an isolated and morphologically obscure example as a parallel would not yet offer strong support for the alleged derivation of *jumala ‘god’. The problem is that the modern Uralic languages do not seem to preserve denominal noun suffixes that could be plausibly connected with this hypothetical PU suffix *-lA; there is a productive oikonym suffix of this shape in Finnic, but it would be semantically and functionally arbitrary to compare the opaque ending in PFi *jumala ‘god’ and PU *ńomala ‘hare’ to a morpheme that forms house and settlement names. Also, there is an unproductive and rather infrequent denominal adjective suffix of the same shape, found in formations such as Fin vetelä ‘watery, washy; limp’ (← vesi : vete- ‘water’). Frog (2020: 215) has suggested that *juma-la could have originated as a derived adjective (‘divine one, one of the sky’), but this is speculation as the word is nowhere attested as an adjective.

Motivated by the obscurity of the alleged “suffix”, Saarikivi (2014: 207) has suggested that PFi *jumala could be analyzed as an obscured compound word instead. He proposes that it developed from earlier *juma-ülä, containing the relational noun root PFi *ülä- ‘place on / above / up / high’ (with widely attested Uralic cognates) as its second member. This explanation fails to convince, however. While it is true that words meaning ‘upper’, ‘high’, and the like, commonly occur as epithets of supreme deities, one expects such an epithet to occur as the modifier of the compound word, not as the head: cf., e.g., Fin yli-jumala ‘supreme deity’ (yli- < *ülä-j-). One would expect a compound noun with the relational noun *ülä as its head to denote some kind of concrete locality, but not an entity like ‘god’. What is more, I have not found any actual examples of Finnic compound nouns that contain the relational noun *ülä as their head, nor of Saami ones containing the cognate relational noun PSaa *e̮lē. Therefore it is highly unlikely that PFi *jumala ‘god’ has developed from such a compound noun.

When the etymology connecting PMari *jŭmǝ and PFi *jumala is evaluated solely by its own merits, the presence of an opaque suffixal element in the latter form could not be seen as a conclusive argument that would refute an otherwise completely regular comparison. The situation changes, however, as an alternative etymology can be proposed for the Mari word.

The new etymology: Mari *jŭmǝ < PU *jilma ‘sky’

I suggest that the generally accepted comparison of Mari *jŭmǝ to Fin jumala ‘god’ is wrong, and that the word actually belongs in the following well-known Uralic cognate set:

• SaaS elmie ‘sky, heaven; air; storm’, SaaU albmie ‘sky, heaven’, SaaP albme, SaaL almme, SaaN albmi, SaaI alme ‘sky, heaven; storm’, SaaSk âʹlmm, SaaK âʹllm, SaaT âʹllme ‘sky, heaven; storm; world’ (< PSaa *e̮lmē)
• Fin ilma ‘air; weather’, Ol ilmu ‘air; weather; world’, Veps iľm ‘air’, Vote ilma ‘air; weather; world’, Est ilm ‘weather; world’, Võro ilm, Liv īlma ‘world’ (< PFi *ilma)
• Komi jen (jenm-) ‘god’, KomiJ jen ‘sky, heaven; god’, Udm in (inm-) ‘sky, heaven’ (< PPerm *jen : *jenm-)
• KhIrt itǝm, KhKaz jeʟǝm, KhO ilǝm ‘world, weather’ (< PKh *ēlǝm)
• MsW jēlǝm, jelǝm, elǝm, MsN ēlǝm ‘weather; time; life’ (< PMs *īlmǝ)

Let us first consider the phonological reconstruction of this word. Both UEW and Sammallahti (1988: 541) reconstruct is as PU *ilma, which agrees with the Saami and Finnic forms; the Finnic cognates indeed quite unambiguously point to an original disharmonic stem with the vowel combination *i–a. The vowel e in Komi and KomiJ jen also agrees with this reconstruction: Komi e goes back to PPerm *e, which is the regular reflex of both PU *i(–a) and *i(–ä). However, the vowel correspondence Komi e ~ Udm i is very rare, and it apparenly occurs in no other word of Proto-Uralic origin, so the Permic evidence as a whole is somewhat ambiguous. A more difficult problem is encountered with the Ob-Ugric forms, which contain a front vowel (PKh *ē, PMs *ī) although the expected reflexes of the PU disharmonic vowel combination *i(–a) are PKh *a / *i̮ and PMs *a.

In addition to the aberrant vowel correspondence there is also another noteworthy phonological feature: the word-initial j- in Komi jen. One can indeed plausibly explain the exceptional vowel correspondence Komi e ~ Udm i by reconstructing the PPerm form as *jen : *jenm-, and assuming that the word-initial palatal glide prevented the otherwise regular development PPerm *e > Udm e, o. The question is, then, where the word-initial *j- comes form. UEW states that it is secondary, but provides no argument or justification for this view. The problem is that the secondary addition of prothetic j- would not be a regular development outside Jazva Komi: compare the correspondence Komi jen ~ KomiJ jen ~ Udm in against that in Komi eski̮- ~ KomiJ jeski- ~ Udm oski̮- ‘believe’. In order to plausibly account for these two different patterns we can reconstruct PPerm *jen : *jenm- ‘sky / god’ with word-initial *j-, in contrast to PPerm *esk- ‘believe’ without word-initial *j-. One should note that this issue is not directly connected with the word-initial j- in some of the Ob-Ugric forms (KhKaz jeʟǝm, MsW jēlǝm, jelǝm): these can indeed be regarded secondary, as maintained by UEW, because the addition of prothetic j- before word-initial *ē- or *ī- is indeed regular in certain Khanty and Mansi varieties.

As the traditional PU reconstruction *ilma does not account for the sound correspondences in this cognate set, it needs to be revised. It is important to note that we are not dealing with a unique irregularity here, as Finno-Saamic *i- ~ PKh *ǟ- / *ē- ~ PMs *ī- is actually a regular sound correspondence (note that in the Khanty system of ablaut, PKh *ē represents the mid ablaut grade of underlying *ǟ (Zhivlov 2007), and the two vowels thus reflect the same Pre-Proto-Khanty vowel). Pystynen (2015) notes that in addition to Fin ilma and its cognates, it also occurs in the following cognate set:

• Fin ien (pl. ikenet), Ol igen (pl. igenet), Veps igeń, Vote ičemet (pl.), Est ige (pl. igemed), Liv i’gmõd (pl.) ‘gum(s)’ (< PFi *igen : *ikene- ~ *ikeme-)
• KhVVj äɣǝṇ, KhSur ȧɣǝṇ, KhIrt ȧŋǝṇ, KhNi aŋǝn, KhKaz aŋǝṇ, KhO ȧŋǝn ‘chin, lower jaw’ (< PKh *ǟɣǝṇ)
• MsT jīn, MsK ǝɣǝn, ēn, MsW jēn, MsN ēŋǝn ‘chin, lower jaw’ (< PMs *īɣnǝ)
• Hung íny ‘gum’

I have myself proposed a third example of the same correspondence (Aikio 2015: 8–10):

• SaaL amás, N amas, I oomâs, Sk õõmâs, K ā̮ma̮s ‘unfamiliar, strange, odd’ (< PSaa *e̮me̮s)
• Fin ihme, Vote ime ‘wonder, miracle; strange’, Est ime, Võro imeh, imeʔ, Liv i’m ‘wonder, miracle’ (< PFi *imeh)
• KhVVj jim, KhNiz jem, KhKaz jεm, KhO jem ‘taboo’ (< PKh *jēm); KhVVj jimǝŋ, KhIrt KhNiz jemǝŋ, KhKaz jεmǝŋ, KhO jemǝŋ ‘sacred’ (< PKh *jēmǝŋ)

Yet a fourth example of the correpondence can be proposed. The following verbs may be cognate and go back to PU *jičV-:

• ? MdE ičems, MdM ičǝms (past.3sg ičś) ‘knead’ (< PMd *ičǝ- : *ič-)
• ? Mari:E jǝ̑čem, MariW jǝčem ‘ache, smart, hurt’ (< PMari *jĭč-(e-))
• ? KhIrt ič- ‘push; press; shove’, KhNi KhKaz ĭš-, KhO is- (passive) ‘get damaged by pressure’ (< PKh *īč-)
• ? MsT jīš-, MsK ješ-, jes-, MsW jēš-, ješ-, MsN jēs- ‘rub; chafe’ (< PMs *jīš-)

UEW presents the last etymology in a different form: the Mansi verb is not mentioned, and another Mari cognate is suggested (MariE išem, MariM iśem, MariU jǝ̑šem ‘press, squeeze, grasp, clasp, clench’ < PMari *is-(e-) ~ *jĭs-(e-)). This is clearly in error: according to Bereczski (2013) this Mari verb was instead borrowed from Early Middle Chuvash *xis- (> Chuvash xĕs- ‘press, squeeze, grasp, clasp, clench’), and the Mari sibilant *s could not reflect PU *č anyway. However, even in its modified form the aforementioned Uralic etymology is not entirely straightforward: the Mansi verb clearly had a word-initial *j- already in Proto-Mansi (in contrast to PMs *īlmǝ and *īɣnǝ), and the vowel of the Khanty form (PKh *īč-) has to be assumed to represent the I-grade of underlying *ǟ, which developed due to the presence of an umlaut trigger in the second syllable which was later lost. But in any case the reconstruction *ičV- proposed by UEW can be ruled out as impossible; it also does not explain PMd *i-, the default development being PU *i > PMd *e.

As there are at least three (and perhaps four) independent examples of the sound correspondence discussed above, it appears obvious that it results form regular phonological development of some sort. The question, then, is what exactly this development was. One can immediately rule out the idea that this exceptional vowel correspondence was conditioned by the occurrence of *i in word-initial position. In Ob-Ugric the default reflexes of PU *i are PKh and PMs *ä, and this development is also attested word-initially:

• PU *imi- ‘‘suck’ > PKh *äm- (> KhVVj em-, KhSur ȧ̆m-, KhIrt em-).
• PU *ipsi ‘smell (noun)’ > PKh *äpǝl (> KhVVj ewǝl, KhSur ȧ̆pǝʟ, KhIrt epǝt, KhNi epǝt, KhKaz epǝʟ, KhO epǝl), PMs *ätǝ (> MsT ät, MsK MsW ǟt, MsN at) . — Note that UEW cites also another Mansi cognate, PMs *(j)īp ‘steam, vapor’ (> MsT jīp, MsK ep, MsW jep, MsN ēp), and reconstructs the word as PU *ipV ~ *ipV-sV ~ *ipV-ćV. This extremely odd reconstruction and the ad-hoc morphological segmentation are motivated by the aforementioned Mansi noun and by MdM opǝś ‘smell, scent’, but both are phonologically so aberrant that there is no reason to assume they have any etymological connection to PU *ipsi in the first place. MdM intervocalic -p- is the regular reflex of the PU geminate stop *-pp-, and moreover, the development PU *i > MdM o would imply a PU disharmonic stem which is not supported by any of the other forms in the cognate set. As regards PMs *(j)īp ‘steam, vapor’, there is absolutely no reason to postulate an etymological connection to PMs *ätǝ ‘smell’ because the gap between these two completely dissimilar Mansi words cannot be bridged by any known phonological developments and derivational processes.
• PU *iskä- ‘believe’ > PKh *äɣǝl- (> KhVVj öɣǝl-, KhSur ȧ̆ɣʷǝʟ-, KhIrt ewǝt-, KhKaz ewǝʟ-, KhO ewǝl-), PMs *äɣt- (> MsK äɣt-, MsN aɣt-). — The only cognates are found in Permic (Komi eski̮-, Udm oski̮-). For unclear reasons UEW reconstructs the verb as *eskV-, and Sammallahti (1988: 543) as *äski-, although the vowel correspondences quite unambiguously point to PU *i–ä.
• PU *itä- ‘appear, come out’ > PKh *ät- (> KhVVj et- ‘wax (of the moon)’, KhSur ȧ̆t- ‘dawn; come out, grow’, KhIrt KhNi et- ‘arise, stand up; dawn; come out, grow’, KhKaz εt-, KhO et- ‘arise, stand up; rise (of the sun)’). — The Khanty verb has cognates in Finnic (cf. Fin itää ‘sprout, germinate’) and Samoyed (cf. Ngan ŋǝtǝďa ‘see’, ŋǝδusi̮ ‘be visible; seem’). Both UEW and SSA regard the comparison between the Finnic and Khanty verbs as uncertain, but without any good reason; the Samoyed cognates discovered by Helimski (2000: 199) further verify the etymology. The meaning of the Finnic verb is a secondarily development; the original meaning is indirectly attested in SaaN ihtit ‘come out, appear’ (< PSaa *itē-), which on account of its first-syllable vowel must have been borrowed from (Proto-)Finnic, and also implied by the related Fin itä ‘east’ (“the direction of sunrise”; cf. Latin oriēns ‘east’ ← orior ‘rise, appear, become visible’).

Thus, the correspondence pattern seen in the Ob-Ugric cognates of Fin ilma, ien and ihme cannot be explained by reconstructing the PU forms with word-initial *i-, and another solution is needed. A clue is provided by the word-initial j- in Komi jen ‘god’. Pystynen (2015) suggests that the word-initial sequence *je- should be reconstructed for the PU forms of Fin ilma and ien, and that a regular development PU *je- > *ji- > *i- took place in Finnic (and probably in Saami, too). In Ob-Ugric the PU sequence *je- would then be reflected as PMs *ī- and PKh *ǟ- / *ē-. Pystynen’s suggestion was earlier supported by me (Aikio 2015: 8–10), but there is an obvious problem: if PU *e is postulated as the original first-syllable vowel, then there is no way to account for the disharmonic vowel combination in PFi *ilma ‘air, weather’. The vowel *e is not known to have been harmonically neutral in Proto-Uralic, and no vowel combination *e–a can thus be reconstructed. For this reason Pystynen postulates the front-harmonic proto-form *jelmä and assumes an ad hoc irregular change *ilmä > *ilma in Proto-Finnic (cf., e.g., Fin silmä ‘eye’ < PFi *silmä < PU *ćilmä, which shows no change in its second-syllable vowel).

Zhivlov (2023) reconstructs PU *jilma ‘sky’ and *jikin : *jikini- ‘gums’, thus modifying Pystynen’s proposal as well as his own earlier reconstruction *ikin (2016: 297, 299), which cannot explain the vowel development in Ob-Ugric, as we saw above. This solution is presumably motivated by the problem discussed above, although Zhivlov does not explicitly mention the issue. This is indeed a much better alternative, as it allows us to assume that the disharmonic vowel combination in PFi *ilma was simply regularly inherited from Proto-Uralic. To explain the Ob-Ugric reflexes, then, Zhivlov needs to make two assumptions: 1) that the word-initial palatal glide prevented the default development of PU *i(–a) into a back vowel, and 2) that the Ob-Ugric reflex of PU *ji- is different from the default reflex of PU *i. The second assumption is unproblematic as it is already supported by two parallels, the cognates of Fin ien and ihme discussed above. I will return to the first assumption below.

A phonological parallel: Mari *jŭt ‘night’ < PU *jita

At this point it is illuminating to look at the Mari noun *jŭt ‘night’, as it provides a perfect phonological parallel for the new Uralic etymology I’m suggesting for Mari *jŭmǝ ‘sky, god’. UEW suggests that the Mari word belongs in a cognate set with Ob-Ugric and Samoyed cognates:

• MariE jüt, MariV jŭt, MariNw jŏt, MariW jǝ̑t ‘night’; MariE jüδǝ̑m, MariV jŭδĭ̮m ~ jŭδŭm, MariNw jŏδŏm, MariW jǝ̑δǝ̑m ‘at night’ (< PMari *jŭt : *jŭdǝ-m)
• KhSur KhIrt ȧt, KhNi KhKaz at, KhO ȧt ‘night’ (< PKh *ǟt); KhVVj KhSur KhIrt itǝn, KhNi KhKaz jetǝn ‘evening; in the evening’ (< PKh *ētǝn)
• MsT jīt, MsK īť, MsW ēť, īť, MsN ɛ̄t, ēť, ēťi ‘evening, night’ (< PMs *īt ~ *īť)
• Slk *ǖtǝ (Ta ǖti̮, Ty ǖdǝ ‘evening’), Kam nu̇ďi, Mat ńǖdɜ, nǖdɜ ‘evening’ (< PSam ? *üǝtV ~ *ńüǝtV)

The Mari noun displays irregular dialectal variation between front and back vocalic forms: the forms in Hill Mari and in the Northwest and Volga dialects are back vocalic and suggest the PMari reconstruction *jŭt : *jŭdǝ-. Evidently this is the most archaic form, and the front vowel in most Meadow and Eastern Mari varieties resulted from an irregular change *u > ü which was influenced by the preceding palatal semivowel j-. There are also a few other words showing the same change: e.g., MariE jük, MariNw MariW juk ‘sound, voice’ (< PMari *juk), MariE jür, MariV MariNw MariW jur ‘rain’ (< PMari *jur), MariE jülem, MariV jŭlem, MariNw jŏlem, MariW jǝ̑lem ‘burn’ (< PMari *jŭl-e-). As can be seen, this irregular change has affected reflexes of both PMari *ŭ and *u, which have merged into u in most Meadow and Eastern varieties.

UEW reconstructs the proto-form as *jitV. However, the etymology cannot be valid in the form presented by UEW, because the dictionary fails to notice that within Mari the word must be reconstructed as back vocalic. Bereczki (2013) points out this fact, and rejects the Uralic etymology altogether for this reason. However, we can still regard the Proto-Mari back vowel *ŭ secondary from the point of view of Proto-Uralic, if we reconstruct a disharmonic stem *jita. This is obviously the correct solution, as the sound correspondence PMari *jŭ- ~ PKh *ǟ- / *ē- ~ PMs *ī- has an exact parallel in the reflexes of PU *jilma ‘sky’.

The proposed Samoyed cognates, however, do not match the PU reconstruction *jita at all, and therefore they must be removed from this cognate set. They point to the Proto-Samoyed rounded front vowel *ü, and moreover the long vowel of the Selkup and Mator forms suggests some kind of original sound sequence instead of a simplex vowel, presumably PSam *-üǝ-; note that reconstructing a sequence *-üj-, *-uj-, or the like, is not possible in this case because PSam *-jt- is reflected as -st- in Mator. Moreover, the alternation between vocalic initium and *ń- is anomalous. If the original form can indeed be reconstructed as PSam *üǝtV, then it is tempting to hypothesize that it is an obscured derivative of the most widespread Uralic word for ‘night’, which has often been reconstructed as PU *üji.

Further parallels for PMari *ŭ < PU *i(–a)

The etymologies for PMari *jŭmǝ ‘sky; god’ and *jŭt ‘night’ presuppose that PMari *ŭ is the regular reflex of PU *i in “disharmonic” stems (i.e., ones containing the vowel combination *i–a). Two more words showing this development were cited by me in my paper on the reconstruction and development of the Proto-Mari vowel system (Aikio 2014: 156):

• PU *kićka- ‘tear’ > PMari *kŭšk-ed-ä- > MariE kuškeδam, MariW kǝ̑škeδäm ‘tear off, tear in two’. — The Mari verb is cognate with SaaN gaikut ‘tear’, Fin kiskoa ‘pull hard, tug’, Komi koś-, Udm keśi̮- ‘tear, rip, peel’, KhVVj kös- ‘tear, tear down, break up’, Hung hasít ‘cleave, split, rip, tear’.
• PU *wiša ‘green’ > PMari *ŭž-ar > MariE užar, MariW ǝ̑žar ~ žar ‘green’. — The Mari word is cognate with Fin viha-nta ‘lush, green (of plants), viherä (obsolete) ~ vihreä ‘green’, MdE ožo ‘yellow, pale’, Komi vež ‘green, yellow’, and Udm vož ‘green’.

In addition, yet another two new examples can be proposed:

• PU *wiša ‘hatred’, *wiša-ta- ‘hate’ > PMari *ŭžed-a- > MariE užeδam ‘hate’. — The Mari verb is cognate with Fin viha ‘hatred’, vihata ‘hate’, Komi vež ‘envy, anger, malice’ and Udm vož ‘hatred, anger’. It is remarkable that this completely straightforward etymology does not appear to have been previously proposed; probably this is because the Mari verb is rare, being attested only in some Eastern dialects.
• PU *kira- ‘anger / curse (?)’ > PMari *kŭrǝ-kt-e- > MariE kurǝktem, MariNw kŏrŏktem ‘lie, blather, babble’. — Cognates include SaaN garru ‘curse (noun)’, garuhit ‘curse, swear’, Fin kiro ‘curse (noun)’, kirota ‘curse, swear’, MdE kor ‘annoyance, anger’, KhVVj kăram- ‘get angry’, MsK xɔr-, kʷår- ‘quarrel’, and Hung harag ‘anger, wrath, rage’. Although the original meaning is somewhat difficult to reconstruct, the semantic differences pose no problem to the etymology: the meaning ‘curse’ can easily develop from the sense of verbal expression of anger or hostility, and moreover, also the meaning ‘abuse (verbally)’ is well-attested in Finnic. As further parallels one can mention MariE jatlem ‘abuse, revile; reprimand, reproach; curse, damn’; OHGerm fluohhōn ‘curse; abuse (verbally)’; Spanish maldecir, Italian maledire ‘curse’ < Latin maledīcēre ‘slander, speak ill of; curse’ (← male ‘badly, wrongly, wickedly’ + dīcēre ‘speak’).

There are thus six etymologies in total that display the vowel development PU *i(–a) > PMari *ŭ; this should easily suffice to prove that the development is regular.

Comparing the two etymologies for Mari *jŭmǝ

Now that we have a two competing etymologies for Mari *jŭmǝ ‘sky; god’, we need to assess which one is likely to be correct. It is crucial to note that also the new comparison of Mari *jŭmǝ to PU *jilma ‘sky’ is phonologically fully regular. As was seen above, a total of six examples prove that PMari *ŭ is the regular reflex of PU *i in disharmonic stems. Moreover, the word-initial *j- in PMari *jŭmǝ is in full accordance with the revised PU reconstruction *jilma that was argued for above, and the word *jŭt ‘night’ (< PU *jita) even offers an exact parallel for the special sound correspondence between Mari and Ob-Ugric that is found only in reflexes of PU stems of the shape *ji–a. Also the loss of the lateral *l in Mari is predictable, as there are two already established examples of this development in exactly the same phonological environment (PMari *ŭ_m):

• MariE kum, MariW kǝ̑m ‘three’ < PMari *kŭm < PU *kulmi / *kolmi
• MariE kumǝž, MariW kǝ̑mǝ̑ž ‘birch-bark’ < PMari *kŭmǝž < PU *kolmiš (Aikio 2013: 168­­­–169)

Thus, we have two hypotheses to compare: PMari *jŭmǝ ‘sky; god’ reflects either PU *jilma ‘sky’ or a distinct noun *juma from which also PFi *jumala ‘god’ was derived. Both etymologies are phonologically and semantically unproblematic, but there are two factors that make the new etymology superior to the established one. First, the alleged connection with PFi *jumala forces us to assume that the Finnic noun contains an unknown and opaque denominal noun suffix *-lA, although otherwise it could be simply regarded as a momomorphemic trisyllabic noun-root. Second, the assumption that PMari *jŭmǝ continues PU *jilma ‘sky’ does not require us to postulate any new etyma on the proto-language level, because the noun *jilma must in any case be reconstructed to Proto-Uralic regardless of the origin of the Mari word. The situation with the noun *juma is different, as its status is wholly dependent on the etymological hypothesis that connects PMari *jŭmǝ with PFi *jumala. If we assume that the Mari word regularly developed from PU *jilma ‘sky’, there is no longer any need to assume that the alleged (Proto-)Uralic noun *juma ‘god’ has ever even existed.

Thus, the etymology presented here provides a more convincing explanation of the origin of Mari *jŭmǝ, and the apparent match with Finnic *jumala can be classified as a “chance correspondence”, a phenomenon I’ve discussed more extensively in this previous post. The only remaining loose end is the origin of PFi *jumala ‘god’, which can now be approached from slightly different premises, as there is no longer a reason to treat the part *-la as a derivational suffix. According to an old etymology proposed by Paasonen (1907) and later rehabilitated by Koivulehto (1999: 228), the alleged noun *juma ‘god’ was borrowed from Proto-Indo-Iranian *dyumā́n : *dyumánt- (> Sanskrit dyumā́n : dyumánt- ‘heavenly, bright, glorious’) or *dyumná- (> Sanskrit dyumná- ‘splendour, (heavenly) glory’); the etymology has been recently supported also by Holopainen (2019: 107–108). The implied sound substitution *dyu- [= *dju-] > *ju- appears possible, even though it has no known parallels, so it is worth considering whether the Indo-Iranian word could indeed account for the origin of PFi *jumala. The problem is that the Indo-Iranian form does not match the ending *-la in Finnic; on the other hand, in its previously proposed form the etymology implied that *-n : *-nt- in the Indo-Iranian stem was deleted when the word was borrowed to Uralic. As an alternative, one can venture the speculation that PFi *jumala has developed irregularly from Pre-PFi *jumana by dissimilation of two nasals. This is an ad hoc hypothesis, of course, but in other regards the Finnic and Indo-Iranian forms are quite close to each other. But whatever the origin of PFi *jumala may be, there are no valid reasons to assume that it is related to PMari *jŭmǝ, despite the superficial similarity.