The Name of Thor and the Transmission of Old Norse poetry

Abstract

The Old Norse name of the thunder god has the monosyllabic form Þórr in the extant manuscripts. However, one Eddic poem, Hymiskviða, and one skaldic poem, Þórsdrápa, have verses which metrically indicate a disyllabic form with a short first syllable, hypothetically restored as *Þóarr, *Þóurr, *Þonarr or *Þunurr. The existence of such a disyllabic form has been dismissed by recent editors who have resorted to conjectural emendations for the line in Þórsdrápa and mostly ignored the problem in Hymiskviða. When the metrical question is examined in detail, it becomes apparent that there is a great deal of evidence for a disyllabic form. Crucially, Hymiskviða is demonstrably a metrically rigorous poem and both its occurrences of the name Þórr are consistent with a disyllabic form and inconsistent with a monosyllabic form. It is implausible that this is a coincidence. In this case, as in many others, Old Norse poems give us information about linguistic forms which had gone out of use long before the extant manuscripts were written. This is important testimony on the transmission and preservation not only of skaldic poetry but also of Eddic poetry.

Introduction

In the nineteenth century, scholars discovered reasons to believe that the Old Norse theonym Þórr once had a disyllabic form, the presence of which could be demonstrated in two mythological poems—the Eddic poem Hymiskviða and the late tenth-century skaldic poem Þórsdrápa. Among the adherents of this view was Finnur Jónsson, the most prolific editor of Old Norse texts in the early twentieth century, who favored reconstructing the disyllabic form as *Þóarr. In his Skjaldedigtning, the standard edition of skaldic poetry for more than a hundred years, the relevant line from Þórsdrápa is printed as follows (Jónsson, 1912–1915, p. 139):

2.2 Gammleið Þóarr skǫmmum

Likewise, the relevant verses from Hymiskviða were printed by Finnur Jónsson in the following form in his 1905 edition of the Eddic poems (Jónsson, 1905, pp. 114–116):

23.2 dáðrakkr Þóarr

28.4 við Þóar senti

The forms *Þóarr (nominative) and *Þóar (accusative) are reconstructions based on metrical considerations, the spelling of the manuscripts in all three cases indicates the normal prose forms Þórr (nominative) and Þór (accusative).

Recent editors of the two poems have rejected the reconstruction of disyllabic forms. In the Íslenzk fornrit edition of the Eddic poems, the Hymiskviða verses in question are printed with monosyllabic forms and with no notes on any alternative interpretation (Kristjánsson. J & Ólason V, 2014, pp. 403–404):

23.2 dáðrakkr Þórr

28.4 við Þór sennti

The same is true for another standard reference, Hans Kuhn’s edition of the Eddic poems (Kuhn, 1962, pp. 92–93). Furthermore, the Kommentar zu den Liedern der Edda devotes 109 pages to Hymiskviða (von See et al., 1997, pp. 253–361) without mentioning the potential disyllabic forms—a situation characteristic of the short shrift given to metrical considerations by most contemporary Eddic scholars (Males, 2022).

In Skaldic Poetry of the Scandinavian Middle Ages, which aims to be the new standard edition of the skaldic corpus, metrical analysis is frequently discussed and used as a tool for textual criticism. The rejection of the reconstructed form of Þórr is here explicit rather than silent. Edith Marold, the editor of Þórsdrápa, calls the restoration of a disyllabic form “unconvincing” (Marold, 2017a, p. 77), citing an analysis by Lindroth (1917, pp. 167–168). She does, however, agree with previous scholars that a syllable is missing in the line as preserved. Her proposal to fix this is to add the preposition fyr by conjecture. She prints the line as follows (Marold, 2017a, p. 77):

gammleið Þórr fyr skǫmmu

The disyllabic form has, thus, been dismissed and rejected from modern editions. I will, however, argue that the case for its existence is compelling and that its presence in the two poems has interesting implications.

Disyllabic Þórr in Þórsdrápa

The poem Þórsdrápa is attributed to Eilífr Goðrúnarson, one of the court poets of jarl Hákon Sigurðarson of Hlaðir who was overlord of Norway in the late tenth century until his death c. 995. The poem is transmitted in the Prose Edda, where it is cited for its mythological information. The three manuscripts preserving its text are the Codex Regius of the Prose Edda (R), the Codex Trajectinus (T) and the Codex Wormianus (W). The diplomatic text of the line in question is as follows in the three manuscripts:

R: gamleið þor skomv

T: gamleid þor scommom

W: gamleið þorr skommum

The line is five syllables in all three manuscripts, insufficient to cover the six metrical positions of the rigid dróttkvætt meter. The only non-orthographic difference between the manuscripts is that T and W have skǫmmum (“for a short while”) whereas R has skǫmmu (“a short while ago”). Edith Marold picks the R reading as the basis for her emendation to fyr skǫmmu (“a short while ago”). This is, however, the less likely reading, both semantically and stemmatologically. It would be unusual for a poet to describe mythological events as happening recently and it is not typical for T and W to have an error in common against R (Boer, 1924; Þorgeirsson, 2017a; van Eeden, 1913).

The strongest evidence against Marold’s reading of the line is, however, not semantic or stemmatological but metrical. The line she reconstructs (“gammleið Þórr fyr skǫmmu”) is unmetrical and without any parallel in Þórsdrápa. Lines beginning with a disyllabic compound word like gammleið are common in the poem and they end in one of two ways. Most commonly, the next word is a disyllabic word with a short first syllable and a line type known as A2k in Sievers’ system is formed. There are 17 cases of this in Marold’s edition of the poem, the first three are as follows:

1.2 fjǫrnets goða at hvetja (with elision in the fourth position)

1.4 lǫgseims faðir heiman

2.6 Gandvíkr Skotum ríkri

(also 5.2, 5.6, 8.8, 12.6, 12.8, 13.6, 13.8, 14.2, 14.8, 16.6, 17.4, 18.4, 19.4, 21.4)

The alternative way to continue a line beginning with a disyllabic compound is to follow it with an unstressed monosyllable and form an E line. There are 9 cases of this in the poem, the first three are as follows:

6.4 hlaupár of ver gaupu

7.6 hlymþél við mǫl glymja

8.2 hallands of sik falla

(also 10.2, 10.8, 13.4, 14.6, 16.8, 19.2)

There is only one line in Marold’s edition which begins with a disyllabic compound and continues differently than described in the two options above:

2.2 gammleið Þórr fyr skǫmmu (by emendation)

This overview shows how unlikely the emended line is. It is a heavy line reminiscent of the skjálfhent (shivering) type which occurs nowhere in Þórsdrápa, is a very rare type to begin with and occurs in odd rather than even lines (Gade, 1995, p. 58). The reconstructed line could also be compared more generally to A lines with a heavy first drop but that sort of line (also rare in this period and found nowhere in Þórsdrápa) normally involves suffixes or compound names rather than novel root compounds like gammleið (Myrvoll, 2014, pp. 239–266).

Marold rejects another unmetrical attempt to emend the line, which is Lindroth’s proposal of “gammaleið Þórr skǫmmu”, where the idea is to attain the needed sixth syllable by changing the root compound gammleið into a genitive compound gammaleið. This yields a blatantly unmetrical line with no parallel in the poem and, as Marold notes, a violation of Craigie’s law.

It may seem needless to dwell on these failed attempts at emendation but they help establish how rigid and unforgiving—and informative—the dróttkvætt meter is. An attempt to add an extra syllable at random will end in disaster. Adding it after Þórr (as Marold did) will yield an unmetrical line, as will adding it after gamm- (as Lindroth did). Adding a syllable before gamm- or after -leið would (among other problems) violate Craigie’s law while adding one after skǫmmu would ruin the necessary cadence. In fact, the only metrical way to add an extra syllable to the preserved line is to conjecture that Þórr is contracted from a disyllabic form such as *Þóarr.

Furthermore, the normal reason for encountering a five-syllable dróttkvætt line in Old Norse manuscripts is the loss of a syllable through historical linguistic developments. Myrvoll (2014, pp. 318–319) tabulates a total of 72 occurrences of archaic hiatus forms preserved in skaldic poetry and, as a rule, they are not represented in the spelling of the manuscripts.¹ I am unaware of any examples where multiple good manuscripts (such as R, T, and W) agree on a five-syllable line without linguistic developments being the obvious explanation. There is, in fact, another example right in Þórsdrápa. Marold prints Þórsdrápa 6.8 as “œstr, þjóðáar fnœstu”, sensibly restoring the archaic form þjóðáar though all three manuscripts agree on the later form þjóðár.

With the above in mind, we would have to seriously consider the possibility of a disyllabic form of Þórr existing in tenth century Old Norse even if Þórsdrápa was the only evidence for it. But, as things turn out, there is confirmation for this in the Eddic poem Hymiskviða which requires a disyllabic form of the name in both cases where it occurs.

Disyllabic Þórr in Hymiskviða

Hymiskviða is preserved both in the Codex Regius of the Eddic Poems (R, GKS 2365 4to) and in its later sister manuscript AM 748 I a 4to (A). In order to evaluate the metrical argument for its disyllabic Þórr forms we need to first review some facts about Eddic metrics. The traditional Sieversian analysis describes normal verses in the common Germanic meter as consisting of four positions, constituting a minimum of four syllables. In Old English poetry there are strong reasons to believe that any verses shorter than four syllables are due to textual corruption (Pascual, 2014, 2017). In Old Norse, the situation is more complicated. Three-syllable verses are a regular feature of the kviðuháttr meter and they also occur from time to time in some fornyrðislag poems. Such verses are an undeniable feature of Sigurðarkviða in skamma and of Rígsþula. But many poems in fornyrðislag have either no such instances or only very few which are best explained as arising by textual corruption. To uncover the metrical preferences of the Hymiskviða poet, I have collected all instances of three-syllable verses which occur in one or both manuscripts. They are eight in all. Three of them are only in R:

7.2 dag þan fram (R)

7.2 dag fraliga (A)

9.5 Er min fri (R)

9.5 Er minn faðir (A)

21.4 vp sen tvá (R)

21.4 vpp sænn eða tva (A)

Another three verses are only found in A:

12.3 sva forðaz (A)

12.3 sva forþa ser (R)

13.5 fram gængv (A)

13.5 fram gengo þeir (R)

31.1 Harðr a knæ (A)

31.1 Harþr reís akné (R)

Finally, there are two three-syllable verses which R and A agree on:

23.2 daþ raccr þor (R)

23.2 dað rakr þórr (A)

23.3 orm eitr fán (R)

23.3 orm æitrfán (A)

We see that in 6 of the 8 instances, only one manuscript has a three-syllable verse while the other witness has a plausible verse with 4 or more syllables.² This leaves 2 instances which the manuscripts agree on and thus have a stronger claim to authenticity.³ One is 23.3 “orm eitrfán” which is trivial to restore to a normal 4-syllable D-line by expanding the accusative eitrfán to the archaic hiatus form *eitrfáan. Then, finally, we have 23.2 “dáðrakkr Þórr” which would become a normal A2k verse if Þórr can be restored to a disyllabic form. Since the poet avoids three-syllable verses everywhere else it would be quite surprising if he had used it just in this verse with the name Þórr.

The name Þórr occurs one other time in Hymiskviða and again it is the case that a disyllabic form fits best with the meter. This time the case is not as immediately obvious since the verse as preserved, 28.4 “við Þór senti”, is of the C1 type and would not be out of place in some Old Norse poems. However, Hymiskviða is composed in a strict variety of fornyrðislag which prohibits both ictuses in the same C verse from being occupied by long stems. Other poems which follow this principle include Erfidrápa by Gísl Illugason and Hǫfuðlausn by Egill Skallagrímsson as well as poems composed in kviðuháttr (Sigurðsson, 2019, pp. 142–144). To further illustrate the principle at play I have sorted the 49 C verses in Hymiskviða into four categories. The most common type has a long stem in the first ictus and a short syllable in the second ictus:

1.3 ok sumbl-samir

1.6 ok á hlaut *sáu (restored from sá)

2.1 Sat berg-búi

(also 4.3, 5.5, 9.5, 11.1, 12.4, 12.6, 13.6, 14.4, 14.7, 17.2, 17.3, 18.2, 19.8, 20.1, 21.1, 24.2, 25.2, 25.3, 26.3, 27.6, 27.9, 28.1, 28.8, 29.1, 30.5, 31.7, 31.4, 36.5, 37.5, 38.1, 38.5, 39.2)

Another permissible option is to use a short stem in the first ictus and a long syllable in the second ictus:

1.1 Ár val-tívar

6.2 þann lǫg-velli

6.3 Ef, vin-r,⁴ vélar

(also 11.5, 26.5, 28.2, 37.7)

It is also permissible for both ictuses to have a short stem:

10.1 En vá-skapaðr

11.3 Nú er son-r kominn

17.6 ef þú hug trúir

(also 20.5, 22.6)

Thus far, we have 47 cases of C verses which follow the restriction against overly heavy verses. And finally we have 2 verses which at first glance break it, with both ictuses occupied by long stems:

28.4 við Þór senti

32.4 úr knjám hrundit

The reader will know the drill by now. An archaic hiatus form can be straightforwardly restored in verse 32.4 with *knéum instead of knjám which makes 32.4 a verse of the common C2 type. This leaves verse 28.4 with a unique structure in the poem, allowing the conclusion that an archaic disyllabic form is also lurking behind the preserved form Þór. Such a form would render the verse a happily normal C2 verse.

Reconstructions of the disyllabic form

It is implausible that it should be a coincidence that on three separate occasions (one in Þórsdrápa, two in Hymiskviða) the manuscripts agree on metrically unique lines with the name Þórr which would become metrically normal with the restoration of a disyllabic form with a short first syllable. Furthermore, we already know that disyllabic forms of this name existed in West-Germanic—i.e. Old English Þunor, Old Saxon Thunar and Old High German Donar. With this in mind we can ask what the disyllabic Old Norse form looked like but we cannot necessarily provide a conclusive answer. The metrical arguments only tell us that such a form existed but do not let us determine its exact history or form.

An obvious possibility is that Old Norse once had a form like *Þonarr < *Þunaraz or *Þunurr < *Þunuraz.⁵ The form *Þonorr⁶ was first proposed by Konráð Gíslason who quotes the line from Þórsdrápa and suggests an explanation in his typically laconic style:

The name þórr once seems to have had the form þonorr or something similar, similarly to e.g. þinorr gollorr jöfurr. Then on seems to have turned into ó ... perhaps without the word immediately ceasing to be a disyllable.⁷

With the more conventional spelling of u in the unstressed syllable we can represent Konráð’s suggested development in the following way while including the option that the unstressed vowel was a:

*Þonurr > *Þóurr > Þórr

*Þonarr⁸ > *Þóarr⁹ > Þórr

The second phase of this development is relatively straightforward. The development from *Þóurr to Þórr would be a normal hiatus development seen in forms such as dó from *dóu or Gró from *Gróu.¹⁰ This is less straightforward with *Þóarr since óa is normally preserved as such in Old Norse—e.g. in klóa or Gróa. The nearest analogues for the proposed change seem to be búandi > bóndi and skúar > skór. While I do not wish to rule out *Þóarr it seems that *Þóurr would be easier to account for and I will use that form in the remainder of the discussion.

There is no difficulty in getting from a hiatus form like *Þóurr to Þórr. What is more challenging to explain is why a form like *Þonurr, or rather *Þunurr (a form with an urr ending would presumably not undergo a-mutation) would lose its n to begin with since intervocalic n is not normally lost in Old Norse.

Noreen (1913, pp. 72–73) proposed that the loss of n happened first in the dative form and then spread by analogy to the other cases. We can assume that *Þunurr or *Þonarr would decline like jǫfurr or hamarr which have the dative forms jǫfri and hamri. In Noreen’s conception, the dative form *Þunri becomes Þóri since n is often lost before other consonants with lengthening, lowering and nasalization of the preceding vowel—cpr. e.g. Old Norse ósk to Old High German wunsc. The form without n then spreads to the nominative, yielding *Þóurr and ultimately Þórr.

Hjalmar Lindroth (1917) objected to Noreen’s analysis on the grounds that it is atypical for a sound development happening in the dative to spread to the rest of the inflectional paradigm.¹¹ Using another theonym as an example, he points out that the dative form Nirði does not affect the vocalism of the nominative Njǫrðr. In order to explain Old Norse Þórr Lindroth proposed a Proto-Germanic *Þunraz which would yield Þórr in Old Norse easily enough. By explaining the disyllabic West-Germanic forms as arising from the addition of svarabhakti vowels, Lindroth believed his theory to satisfactorily account for all the Germanic facts. He conceded, however, that by rejecting the traditional reconstruction *Þunaraz in favor of *Þunraz some similarity with the Celtic theonym *Tonaros is lost. Indeed, comparativists often prefer *Þunaraz (Birkhan, 1970, pp. 316–332; West, 2007, p. 249).

Both Jónsson (1921, p. 302) and Genzmer (1934, p. 70) responded to Lindroth’s work by briefly reiterating the strength of the metrical arguments for a disyllabic form but without giving any detailed alternative account for the development of the theonym. Finnur’s faith may nevertheless have been shaken since in his 1932 edition of the Eddic poems he prints dáðrakkr Þórr and við Þór senti in Hymiskviða even while systematically restituting hiatus forms such as sǫ́u and ęitrfáan (Jónsson, 1932, 91–99).

The monosyllabic form

Having spent some time pondering the disyllabic form it is time to ask what we can establish about use of the monosyllabic form. For this, there is more metrical evidence to consider. We have three cases where the metrics indicate a disyllabic form but to put this in context we should ask how much metrical evidence we have for the monosyllabic form. The answer may come as a surprise. Among the mythological poems of the Codex Regius, Hymiskviða is a rarity—a metrically rigorous poem which allows us to recover phonological information. Many of the mythological poems, however, use ljóðaháttr which is a freer form where it is generally impossible to distinguish a monosyllable from a light disyllable. There are five occurrences of Þórr in Grímnismál, Lokasenna and Alvíssmál but as far as I can tell there is no way to determine whether they were intended as monosyllabic or disyllabic. The metrical form of Hárbarðsljóð is freer still and perhaps best characterized as alliterative free verse. There are 12 occurences of Þórr there but no way to tell what form or forms the poet used. While Þrymskviða has a somewhat more regular structure, none of the five verses in it in which the name Þórr occurs are regular fornyrðislag verses, whether we assume a monosyllabic or a light disyllabic form. Finally, Vǫluspá has one occurrence of Þórr in the somewhat unusual verse, “Þórr einn þar vá”.¹² This is not easy to categorize but in general if the first lift of a verse is at its beginning it can consist either of a monosyllable or, by resolution, a light disyllable.

The result, then, is that in the Eddic poems of the Codex Regius, there are only two cases where there are strong metrical arguments for determining the form of the theonym—both of these are in Hymiskviða and both of them indicate a disyllabic form.

On the other hand, skaldic poetry of the tenth century offers us some metrically confirmed monosyllabic forms. The Prose Edda cites the poem Haustlǫng by Þjóðólfr ór Hvini (ca. 900) which has the line “þá varð Þórs of rúni” (Clunies Ross, 2017, p. 443). Since Old Norse poetry does not allow resolution on the second ictus, only a monosyllabic form of Þórs is permissible here. Another monosyllabic Þórr form quoted in the Edda is attributed to Kormákr (mid- to late-tenth-century): “heið; sitr Þórr í reiðu” (Marold, 2017b, p. 281). Finally, Þórsdrápa itself has an example of monosyllabic Þórs in lines 11.7 “skalfa Þórs né Þjalfa” and 22.3 “skelfra Þórs né Þjalfa” (Marold, 2017a, pp. 99, 123). More examples exist but I emphasize these because they are attributed to early skalds and the poetry cited in the Edda carries every sign of being authentic unlike the sometimes suspicous lausavísur stanzas in the sagas of Icelanders (see e.g. Males, 2020, pp. 212–276).

A new proposal

I will, finally, sketch out a possible development of the theonym which I believe fits with the data. I do not claim this as the only possibility and if flaws are found in my specific proposal that would not negate the metrical evidence which prompted the investigation.

One possibly relevant piece of evidence is a runic inscription discovered in Hallbjäns in Sweden in 1980. The inscription is dated to ca. 700 and contains the sequence þunurþurus. The tantalizing possibility of reading þunur as a name of Thor has been suggested but the form lacks enough context to establish the matter (Gustavson & Snædal Brink, 1981, 190; Birkmann, 1995, pp. 227–229).¹³ Nevertheless, the fact that we do have one case of runic þunur but no case of þunar may give some tentative support to the reconstruction of *Þunurr.

In Konráð Gíslason’s account, the reconstructed *Þonurr is presented as parallel to the word þinurr (“wood”). Curiously enough, this latter word is preserved dialectically as a monosyllable in Swedish, tjur. In Elfdalian, the vowel is nasalized (tiųor), thus a remnant of the original intervocalic n is preserved (Kroonen, 2013, 538). The þinurr/tjur doublet seems to be a parallel case to the proposed *Þunurr/Þórr situation. Worth mentioning is that Elfdalian also has a nasal vowel in a reflex of Thor’s name, namely in the word tųosdag (“Thursday”).

Hjalmar Lindroth was skeptical that a form arising by sound change in the dative would spread to the rest of the inflectional paradigm. He did not, however, note that the name of Thor may also have been used in the sense “thunder”, where plural usage would be appropriate. This was the case in Old English where þunras or þuneras is the plural nominative of “thunder”. In Old Norse, the entire plural paradigm would have had the conditions necessary for the sound change:

	singular	plural
nom.	*þunurr	*þunraR > *þóraR
acc.	*þunur	*þunra > *þóra
dat.	*þunri > þóri	*þunrum > *þórum
gen.	*þunurs	*þunra > *þóra

There is no difficulty in finding examples where a plural allomorph spreads to the singular by analogy, cpr. e.g. the Old Norse singular gler (“glass”) from the Germanic plural *glazō or the Icelandic singular gæs (“goose”) from the Old Norse plural gæss (singular gás).

The Thor/thunder paradigm would look as follows after analogical levelling has taken place and nasal ó has replaced un throughout:

	singular	plural
nom.	*þóurr	*þóraR
acc.	*þóur	*þóra
dat.	þóri	*þórum
gen.	*þóurs	*þóra

On this model, the disyllabic forms *Þóurr and *Þóur were the forms used in Hymiskviða and Þórsdrápa. They would naturally develop into Þórr and Þór by contraction during the oral transmission and be recorded as such in the thirteenth-century manuscripts.

However, if Þórr is a contracted form derived from *Þóurr it is somewhat surprising to see it so strongly evidenced in tenth-century skaldic poetry, where hiatus forms dominate over contracted forms (Myrvoll, 2014, pp. 309–328).¹⁴ This requires some further explanation. For many hiatus forms, the vowels that are contracted away are recognizable inflectional endings and analogical pressure would have helped maintain them. In the case of *Þóurr, analogical pressure would have worked the other way. After contraction of óu to ó, the word declines according to a common a-stem paradigm, like geirr or bjórr. Thus, the monosyllabic Þórr in early skaldic poetry may be partially or entirely caused by a second round of analogy. The outcome of the earlier analogical development, *Þóurr, would have then existed alongside the monosyllabic form for some time.

Conclusion

The skaldic chronology indicates that the monosyllabic form of Þórr existed as early as ca. 900 and this is supported by tenth-century Danish runestones with the form þur.¹⁵ The presence of the disyllabic form in Þórsdrápa, however, indicates that an alternative archaic form could still be used at the end of the tenth century. This chronological pattern is essentially identical to that shown by another poetic archaism—alliteration with initial vr in words such as vreiðr. The v was lost quite early in West Norse and alliteration confirms the presence of forms such as reiðr in the earliest skaldic poetry. Forms requiring initial vr nevertheless occur in the Eddic poems and in three skaldic poems, the chronologically last of which is Þórsdrápa (Þorgeirsson, 2017b). It is easy to imagine the author, Eilífr Goðrúnarson, delighting in the use of forms from the past. His famously cryptic and difficult poem certainly suggests a poet fond of the obscure (Patria, 2020, pp. 68–76).

Knowledge of metrics and historical linguistics allows us to reconstruct archaic forms such as þjóðáar, knéum, vreiðr and disyllabic Þórr even though none of these are present in the manuscripts of Þórsdrápa or of Hymiskviða. One might have thought that memorizers and copyists of poems would alter lines once historical developments had rendered them unmetrical. However, Leonard Neidorf has concluded that in the case of Beowulf the main class of alterations made in the transmission of the work was the modernization of individual linguistic forms, typically “without regard for their metrical consequences” (Neidorf, 2017, p. 38). The same may be true for many Old Norse poems.¹⁶