The Dating of Hávamál

. Nonetheless, content-based criteria can at


Introduction
The dating of Hávamál has considerable implications for the study of Old Norse culture and literature.If, as some scholars have argued, it was partly or fully composed in the twelfth century, its value as a source to pre-Christian perceptions diminishes greatly.By contrast, if it can be shown to belong to the predominantly pagan period and even to express pagan sentiments, its rich contents are an invaluable source to pagan perceptions, or at least to perceptions among some groups within pagan society.
The aim of the present study is to maximise the number of dating criteria susceptible to some degree of testing, without restricting the eva-luation to linguistic criteria.While such criteria are generally seen as more reliable and susceptible to testing than content-based ones, this need not always be the case.If the criteria are sufficiently specific, it is often possible to come up with some way to test their diagnostic validity.The main difference between linguistic and content-based criteria then becomes that the scholar may need to devise a new test for each contentbased criterion, and there is no way of predicting which parameters may present themselves.By means of this approach, I have singled out six criteria for evaluation, two linguistic and four content-based ones.All of them point to an early date of composition, and I have found no conflicting evidence.
Attempts at dating Hávamál can roughly be divided into linguistic and literary approaches.Linguists typically treat Hávamál among other poems, whereas literary scholars tend to discuss Hávamál on its own.Most literary or content-based attempts at dating Hávamál are epistemologically problematic, since the diagnostic significance of the criteria remains unclear.Thus, Klaus von See and Hermann Pálsson note general similarities to European gnomic literature while disregarding formal criteria (von See 1972a and1972b;Hermann Pálsson 1990).John McKinnell argues for a twelfth-century interpolation of stanzas 84 and 91-110, drawing on the erotic works of Ovid.He takes formal criteria into account and notes that the evidence, although limited, suggests that these stanzas are old (McKinnell 2005: 92-93). 1 He therefore argues that a twelfth-century poet drew on older poetic material when composing stanzas 84 and 91-110 (McKinnell 2005: 100).
All three scholars focus on textual analogues that relate to widely shared aspects of the human experience.While none of the features mentioned are typologically rare or specific enough to suggest influence, noteworthy differences are passed over in silence.Thus, as David Evans points out, von See (1972a and1972b) assumes influence from Hugsvinnsmál on Hávamál but does not account for the fact that while Hugsvinnsmál contains many references to Christian faith and book culture, Hávamál has nothing of the kind, and while Hugsvinnsmál forbids sac-1.The formal criterion suggesting that these stanzas are old is the frequency of expletive of.By contrast, McKinnell assumes that alliteration in r-in the name Rati means that stanza 106 must have been composed after c. 900, after an initial v-had been lost.It is not known whether this name ever had initial v-, however ( de Vries 1962: s.v.).
rifice, Hávamál gives instructions on how best to perform it (cf. Evans 1986: 15-18).In other words, not only does von See focus on similarities that are too vague to allow for evaluation of their diagnostic significance, but he also ignores differences that might have served to test the plausibility of his claim.Unlike von See, Hermann Pálsson (1990) does not delimit the hypothetical influence to any given text, but takes any analogues he can find as diagnostic of influence.He thus employs a method guaranteed to produce the desired results, and like von See, he does not take note of differences that might have served to test his hypothesis (Evans 1990(Evans -1993)).
McKinnell (2005) draws on both von See and Hermann Pálsson and notes comparable sentiments in Hávamál and Boethius, Ovid's Ars amatoria and Vergil's Eclogues.Like in previous examples, the analogues consist in widespread motifs, in this instance the mutual deviousness of the sexes in erotic affairs.The case of Ars amatoria is instructive.While Hávamál places its main emphasis on women's deceptiveness against men, Ars amatoria explicitly says that this is rare and that men are the main deceivers (see quotations in McKinnell 2005: 97-98).The two texts are thus in disagreement, a problem that McKinnell avoids by suggesting that the influence of Ars amatoria starts after Hávamál's treatment of men.Within McKinnell's Hávamál B (stanzas 84 and 91-110), however, focus shifts several times between women and men, and dissecting such a thematic unit in order to compare only one aspect of it to another text runs the risk of producing a likeness that is subsequently taken to be diagnostic.For various reasons, then, these content-based attempts at dating all or parts of Hávamál fail to convince.
Other scholars have argued that Óðinn's hanging in stanzas 138-45 is influenced by the Christ's crucifixion (Lassen 2009;Males 2013: 106-12).Some of the analogues are indeed so specific as to suggest direct influence, and Lassen and the present author have both assumed that this part of Hávamál must therefore have been composed in the Christian period.2This is a questionable conclusion, however, since there are many indications of influence from Christian cultures before official Conversion, including influence on mythological motifs and perhaps even ritual (Marold 1974;Males 2022).
A variety of the "Christian" line of argumentation was presented by Klaus von See. In 1987-89, von See launched into a fierce diatribe against David Evans' edition of Hávamál (1986).Overall, von See's arguments are either inconsequential or ad personam, claiming that Evans is a dilettante, unacquainted with "historisch-philologischen Argumentationsmetode", etc. (von See 1989: 142).Von See's own example of what constitutes good "historical-philological argumentation" demands that the reader accept that Hávamál 79 must be dependent on Hugsvinnsmál 73, even though the only shared lexical item is munúð 'love', and although Hávamál 79 warns us not to be stupid, whereas Hugsvinnsmál 73 admonishes us to avoid sin (von See 1987: 139-40).As Evans notes, with such a definition of good method, von See can corroborate any claim he wishes, and the procedure is further aided by von See's assumption that words found in Christian texts were only ever used in such texts.This assumption can easily be falsified by recourse to, e.g., Greek and Latin literature (Evans 1989: 131-37).Evans' insistence that probabilities be weighed against each other without bias was emphatically rejected by von See and was not taken into account in the subsequent work on the Kommentar zu den Liedern der Edda (1997Edda ( -2019)).Indeed, Evans' observations on the importance of sound method have had little impact on eddic scholarship overall.
As far as I can tell, only one of von See's arguments was not sufficiently dealt with by Evans, namely the claim that the name Hávi 'the Tall/High One' for Óðinn is Christian and therefore late.The name occurs only in Gylfaginning and Hávamál (stanzas 109, 111, 164;von See 1989: 146-47).Von See suggests that the name has been extracted from Gylfaginning, but it occurs there only once, in an allusion to Hávamál, which has just been quoted (von See 1987: 146;Snorri Sturluson, Prologue and 'Gylfaginning': 8).In accordance with his overall method, Snorri here alludes to his poetic source (Males 2021).In addition, von See apparently takes 'the Tall/High One' to mean 'exalted' or even 'the one in heaven'.Based on descriptions of Óðinn as a tall man, however, we may conclude that it most likely means 'the tall one' (Falk 1924: 15).Von See thus appears to be wrong on two accounts, and to this may be added the general observation that when a deity is known by no less than 169 names, some of these will necessarily be reminiscent of features known from the Christian tradition (Falk 1924: 3-34).Some content-based arguments presented in earlier scholarship command more credence, since they are susceptible to testing.Evans mentions cremation, which will be treated in some detail below, as well as Hávamál 72.4-6 sjáldan standa bautarsteinar | brautu naer | nema reisi niðr at nið ('rarely do bautarsteinar stand close to road unless a relative raise [them] for a relative').Such commemorative stones are unknown in Iceland and belong mainly to the pre-Christian period in Norway (Evans 1989: 129).Stanza 72 is thus suggestive both of an early date and a Norwegian point of origin.
As seen from these observations, few attempts at content-based dating of all or parts of Hávamál command credence, either because of methodological flaws or due to unwarranted conclusions, although there are exceptions.Linguistic criteria generally hold a better claim, since their validity can be tested along a number of parameters.Most notably, the findings can be compared to the datable skaldic corpus.Before turning to linguistic criteria, however, I will briefly treat challenges to their validity.
Bjarne Fidjestøl's The Dating of Eddic Poetry (1999) was a landmark in the discussion of dating criteria, and its methodology is overall sound.Even though Fidjestøl's results were promising, he was extremely cautious in his conclusions.Scholars have responded in different ways to Fidjestøl's study, some focusing on his results, others on his expressions of caution.Haukur Þorgeirsson has performed additional testing on the criteria of expletive of and alliteration in vr-by studying their correlation with other criteria (Haukur Þorgeirsson 2012 and2016).The present author has made a detailed study of the distribution of mythological references in skaldic poetry, showing that Fidjestøl's results are skewed, and that contrary to Fidjestøl's conclusion, the criterion is most likely valid (Males 2020: 39-93).Some of the most important dating criteria therefore appear more reliable today than they did in Fidjestøl's study. 33. Sapp 2022 is a book-length study of linguistic and metrical dating criteria in eddic poetry, and as such, it represents a counterweight to some of the trends so far discussed.Even so, I have opted not to include it in the discussion of advances in the wake of Fidjestøl's study.Sapp adds three new criteria to the ones discussed in previous scholarship: the frequency of sá er, A-lines and lines with heavy dips.The latter two present no clear trajectory, however, and sá er yields results suggesting that its distribution is not chronological.Sapp's new criteria therefore do not appear to be very useful, a topic which I intend to explore in greater depth elsewhere.

The Dating of Hávamál
Other scholars have proceeded further than Fidjestøl in an agnostic direction.Bernt Øyvind Thorvaldsen's chapter on dating in A Handbook of Eddic Poetry (2016) is a case in point.Thorvaldsen discusses Haukur's study of expletive of and violations of the V2-principle (Haukur Þorgeirsson 2012).He notes that the method of correlating two independent and highly frequent criteria is sound, and that the results indicate that the language of some poems is more archaic than that of others.On this basis, however, he draws the following conclusion: Yet even if we accept that these phenomena belong to an early stage of language development, it is far from obvious that the poems themselves do.These features may be preserved within a register that is distinct from everyday language.This clearly renders conclusions drawn from the statistics of occurrences (for example, of the expletive particle) more uncertain.(Thorvaldsen 2016: 80) From the remainder of the article, it becomes clear that by "more uncertain", Thorvaldsen in practice means that scholars cannot rely on such criteria at all (see especially Thorvaldsen 2016: 90).He does not address a point that would have been crucial for the validity of his argument, however.Thorvaldsen refers to the difference between the register of everyday language and that of poetry, but in a study of fornyrðislag poetry only, it is reasonable to assume that the various poems belong to the same register.None of these poems are composed in "everyday language", and even within the poetic corpus, poems in fornyrðislag display about as much linguistic and stylistic "sameness" as one can hope for.Furthermore, Haukur's study is premised on a more fine-grained distinction between registers than that proposed by Thorvaldsen, since Haukur draws attention to the differences between prose, dróttkvaett and fornyrðislag (Haukur Þorgeirsson 2016: 266).Haukur's methodology thus anticipates Thorvaldsen's objections, and the question remains: Why would some poems composed in the same register display more archaic features than others?The most obvious answer is that they are older than the others, and this assumption may be tested against the datable skaldic record.There we see a convergence of mutually independent early features (e.g.vr-, expletive of before noun, hiatus, etc.) in early poets and a corresponding absence in later ones.Thorvaldsen's claim is thus questionable on its own, and in the Old Norse tradition, where we are so lucky as to have the skaldic "control" corpus, it can also be shown to be unlikely on empirical grounds.
Centrally placed in A Handbook of Eddic Poetry, Thorvaldsen's chapter has served to accentuate Fidjestøl's more moderate and nuanced skepticism.Something similar may be said of the Kommentar zu den Liedern der Edda (1997Edda ( -2019)), comprising some 6700 pages and now serving as a point of reference for eddic scholars in general.A recent study has shown that the authors lack an apt metrical analysis, disregard both metrical and linguistic dating criteria and are biased against early dates (Males 2022).Taking a more extreme position than Thorvaldsen, the authors often fail to point out early linguistic forms at all, and when they do, they do not accept even the possibility that such forms may be authentic.Since A Handbook and Kommentar are natural points of reference within eddic studies, the unwary scholar might gain the impression that dating by means of linguistic and metrical criteria is not a viable enterprise.In spite of appearances, however, this impression is not related to the enterprise itself, but rather to the fact that these scholars do not recognise any likely correlation between early linguistic forms and the date of composition of the texts in which these forms are found.Since the skaldic corpus allows us to test this assumption and to conclude that such a correlation exists, the authors' approach amounts to a rejection of probabilities.By contrast, the present article aims to increase our ability to weigh probabilities by evaluating old criteria and identifying new ones susceptible to testing.I turn now to the individual criteria.

Formal criterion 1: vr-
The first formal criterion is alliteration in vr-, which appears to have been replaced by r-by c. 1000 at the latest.Hávamál 32 contains such alliteration in the pair virði : vrekask (later rekask).This criterion has been the object of debate and calls for some elaboration.In the following, I take my starting point in Haukur Þorgeirsson's continuation of Fidjestøl's work, adding observations of my own (Haukur Þorgeirsson 2016).In one regard, Fidjestøl's study is more comprehensive than Haukur's, since Fidjestøl notes that alliteration in r-is attested from the beginning of the skaldic tradition onwards, an observation that Haukur omits (Fidjestøl 1999: 234-35).This means that forms in r-must be treated with The Dating of Hávamál caution for dating purposes, but this is a moot point in the present context, where focus lies on forms in vr-.An important addition to Fidjestøl's study is that Haukur investigates scribal behaviour, and he shows that scribes were unable to supply or correctly analyse vr-alliteration: the main exception is when they had the aid of Óláfr Þórðarson's description of this phenomenon, which he calls vinðandin forna ('the old wynn'), in the Third Grammatical Treatise.In addition, in one out of the three instances of alliteration in vr-in Skáldskaparmál, the archetype clearly had <vr>.The picture that emerges from Haukur's study is more nuanced than the one given in earlier scholarship: scribes apparently lacked the necessary competence for archaising by use of vr-, and this instils confidence in the criterion.A person like Óláfr could have done so, but his description suggests that he was engaging in informed guesswork rather than drawing on tradition.This is a crucial point for the evaluation of whether poets after c. 1000 could have used vr-, and I will therefore explore some neglected aspects of the relevant passage, which runs: […] sem þá at v sé af tekit í þessu nafni vrungu, því at þýðerskir men ok danskir hafa v fyrir r í þessu nafni ok mǫrgum ǫðrum, ok þat hyggjum vér fornt mál vera, en nú er þat kallat vinðandin forna í skáldskap, því at þat er nú ekki haft í norroenu máli.(Den tredje og fjaerde grammatiske avhandling i Snorres Edda: 87) […] as when v is removed in this word vrungu, since Germans and Danes have v in front of r in this word and many others, and we believe that this is old speech, but it is now called 'the old wynn' in poetry, since it is not now present in Norse speech.
One notes here that Óláfr does not express himself like the First Grammarian, who discusses the hiatus form éarn ('iron'), which was being supplanted by járn in his day.The First Grammarian says that 'wise men' have pronounced the word éarn in poetry and that they also attest to having heard others pronounce it that way (The First Grammatical Treatise: 226).By contrast, Óláfr refers to Danes and Germans and says that he "believes" this to be an archaism.This is very different not only from the First Grammarian, but also from Snorri's Edda and other claims about skaldic poetry in the Third Grammatical Treatise itself, which are presented as facts.
Óláfr's use of the word hyggja 'believe' is important for the evaluation of his statement.Overall, grammarians such as Óláfr, Snorri and the First Grammarian speak with the voice of authority and knowledge, which is not conducive to expressions of doubt such as that conveyed by hyggja.Thus, hyggja is never used by the authorial voice in the prose of Snorri's Edda, but only by protagonists, and it is consistently associated with a lack of means to verify the belief.In the First Grammatical Treatise, it is not used at all.In the Third Grammatical Treatise, there is only one occurrence beyond the one quoted above, namely when Óláfr says that the rune u er því first skipat at þat er fremst ok naest sjálfu efni raddarinnar, er vér hyggjum at loptit megi kalla ('is first [in the futhark] because it is foremost [in the mouth] and closest to the very substance of the voice, which we believe may be called "air"') (Den tredje og fjaerde grammatiske avhandling i Snorres Edda: 43).Óláfr is here intent on promoting the futhark as equally sophisticated as the alphabet, and it is apparently for this reason that he presents a natural-philosophical reason for why u is the first vowel in the futhark (on Óláfr's promotion of the runes, see Males 2020: 181-87).
Hyggja is thus appropriate for philosophical speculation, much as it is for religious speculation in Snorri's Edda.This, in tandem with the mention of contemporary Danish and German pronunciation, suggests that the conclusion that vr-is "old speech" is Óláfr's own, presumably developed during his stay in Denmark.This is further supported by the fact that Óláfr first presents another explanation for vr-in vrungu, namely that it is added to rungu: hér er viðlagt 'v' í þessu nafni 'vrungu', til þess at réttir sé stuðlar í dróttkvaeðum haetti ('here v is added in this word "vrungu", so that the staves will be right in dróttkvaett metre') (Den tredje og fjaerde grammatiske avhandling i Snorres Edda: 43).Only after this does he present the opposite explanation, which he "believes" to be correct.
All aspects of this description thus support that the assumption that vr-is old is Óláfr's own, except that he presents the phenomenon as if it were traditionally designated by its own name, vinðandin forna.A number of factors suggest that this is a conceit, however.First, the presentation of a seemingly traditional name probably has little bearing on whether this is in fact the case, since Icelandic grammarians were prone to present such terms as pre-existent in order to achieve a normative style.Thus, for instance, Snorri's nykrat does not conform to the skaldic tradition and seems rather to be based on Horace, and yet the term is presented as if it were traditional.When Óláfr adds the alternative term finngálknat, bringing the terminology even closer in line with Horace's description, he opts for the same strategy (Males 2020: 121-24).Similarly, the Fifth Grammarian uses the seemingly traditional skarbrot ('seam-break') as a calque on Latin hiatus (Males 2016: 128-32).The fact that these authors needed to coin many neologisms is hardly surprising, since before Snorri's Háttatal, there appear to have been no texts treating such technical aspects of poetry.
As seen from these examples, the term vinðandin forna cannot be taken as positive evidence of a longer prehistory, but this does not itself make such a prehistory unlikely.Other factors do, however.First, if there were such an established term, there would be no need for Óláfr to express caution in his guess, since the name itself would then provide him with the authority he needed.Instead, he first presents his arguments and then a name that presupposes them.Second, vinðandin is apparently named after venð, most likely modelled on Old English wynn, designating the rune and letter Ƿ that with some adaptation was borrowed from Old English into Old Norse.The name venð is only used in the Third Grammatical Treatise itself, however, where it designates consonantal /v/.This is apparently an innovation, since the First Grammarian conflated wynn and y and thought them both superfluous (The First Grammatical Treatise: 238).Óláfr's introduction of venð improved the First Grammarian's analysis considerably, and his innovation indicates that neither venð nor vinðandin forna were traditional.All in all, a considerable amount of evidence indicates that vr-was not commonly known to be an archaic feature, and this is supported by the fact that except for the passage quoted above, where scribes were guided by Óláfr's description, all scribes but one failed to produce forms with vr-.
The trace of <vr> in Skáldskaparmál is important to flesh out the picture.It is found in a quotation of a half stanza by Bragi which most likely derives from the archetype, even though it is missing in the manuscript U. 4 The manuscripts R and T have ravngum, A has vravngum, 4. The section on ókennd heiti has seen considerable rearrangement in ABU, and it seems likely that the section on heiti for the sea, where Bragi's half stanza belongs, has fallen out in U in the process.In RTC, it is found between heiti for birds of the battlefield and heiti for fire, which follow in sequence in U, and in AB it is found before heiti for fire, as in RTC.AB are closely related to U. See the overviews of content B has orǫngum and C vaungum.The form in A is historically correct, and the one in C clearly goes back to the same form.The B scribe appears to have interpreted <vr> as the negation ú + r-and then to have changed this "Norwegian" spelling of the negation to Icelandic <o> (= ó-) (Haukur Þorgeirsson 2016: 50-52).The scribe of the earliest exemplar must have understood the form correctly, however, since only then does the spelling reflect alliteration.Curiously, there is no trace of vr-in the other two instances in Skáldskaparmál where the metre requires it: in a half stanza by Eilífr and at one point in Haustlǫng.The first of these would likely also have been present in the archetype, and it is therefore an open question whether Snorri himself restored vr-in one instance but not in another, or if the spelling in Bragi might rather be due to the use of a written exemplar of Bragi's Ragnarsdrápa for the compilation of Skáldskaparmál.The latter explanation is made plausible by indications that Snorri drew on written exemplars of Vǫluspá and Þjóðolfr Arnórsson's Sexstefja, and probably also Grímnismál, in compiling Gylfaginning and Skáldskaparmál. 5In any event, the scribe responsible for the reading of the archetype can be shown to have understood the phenomenon of vr-, although others did not (with the possible exception of the A scribe).
We here have two persons -Óláfr and the hand responsible for the spelling of one half stanza in Skáldskaparmál -who could presumably have archaised using vr-had they wished to.This suggests that a few select people might have been able to do so also in the twelfth century, and that they could thus have produced poems exhibiting a feature that thirteenth-century compilers and modern scholars might take as a sign of high age.The question then becomes whether they in fact did.
There are no instances of vr-in skaldic poetry after c. 1000, even in poetry where we see active attempts at archaisation (e.g., with expletive of and metrical types B and C in even lines).In addition, Haukur has noted a correlation between vr-and a high ratio expletive of in eddic poetry, versus r-and a low ratio of the same (Haukur Þorgeirsson 2016: in Snorri Sturluson, Edda.Skáldskaparmál: x, xlix-l; for C, see Edda Snorra Sturlusonar, II: 596-602. 5. See Mårtensson and Heimir Pálsson 2008.It is also possible that traces of vr-have disappeared in all manuscripts in Eilífr's stanza, but this seems unlikely, given that traces are present in three out of five manuscripts in Bragi's stanza and that both main stemmatic branches are represented in Eilífr's stanza (found in R, T, W and U, the last of these representing the U-branch).
The Dating of Hávamál 54).There are thus two mutually independent counterindications to archaization with vr-, to which may be added Óláfr's guesswork.As far as I can tell, the only cause for reasonable doubt might be the pair vreiðr : vega ('angry : kill'), which is common enough that it could have functioned as a formula of sorts (Haukur Þorgeirsson 2016: 35; the pair occurs three times in Lokasenna, three in Fáfnismál and once in Sigrdrífumál).
There are no indications that vreiðr : vega was used after the loss of v-, however, and it may be unlikely that verses that would have been faulty from a synchronic perspective would have been productive.The criterion thus appears to be strong, especially when dealing with unique pairs as opposed to vreiðr : vega.The unique collocation virði : vrekask in Hávamál 32 thus instils the strongest possible confidence in the diagnostic validity of the criterion for dating the stanza to before c. 1000.

Formal criterion 2: expletive of
Hans Kuhn noted that the particle of seems to have a background in lost Germanic prefixes and that it displays a lowering frequency over time (Kuhn 1929).With his usual reservations, Fidjestøl on the whole accepted Kuhn's results (1999: 207-30).Haukur Þorgeirsson also noted a correlation between vr-and expletive of, and in another study, he observed a correlation between expletive of and violations of the syntactical V2principle, which is another likely archaic feature (Haukur Þorgeirsson 2012;Haukur Þorgeirsson 2016: 54).Hávamál contains a high ratio of expletive of, and both Haukur's studies therefore lend additional support to the poem's relatively high age.There is reason to be cautious, however, since there are a few examples of poets archaising by use of expletive of (Kuhn 1929: 33;Males 2020: 247).
Haukur did not explore the use of expletive of before nouns and adjectives specifically.Kuhn found only 41 examples of the particle before nouns and adjectives in the entire corpus, and in all these examples, the particle plausibly corresponds to lost Germanic prefixes (Kuhn 1929: 25-26).Only two skaldic examples are from after c. 1030, making this a strong dating criterion to before that date. 6Such forms are found in Há-6.I have excluded here the occurrence in Einarr Skúlason's lausavísa 4.2, where the interpretation of of as the expletive particle involves an emendation which is rejected vamál stanzas 4.4 um oeðis; 21.6 um mál; 38.6 um þǫrf. 7These stanzas thus have almost as strong a claim to dating before the early eleventh century as stanza 32, containing the form in vr-.
Formal criteria thus unambiguously point to a date before c. 1000, but the evidence is somewhat underwhelming.Only for four stanzas do we have strong evidence, and these are all found in the the first quarter of the poem.Since Hávamál falls into several thematic sections, this raises the question of whether the testimony of these criteria is valid for the poem at large.Additional evidence would therefore be highly desirable, and for this, we must turn to content.As noted above, attempts at dating Hávamál by means of content have for the most part disregarded formal criteria, as well as specificity generally, and they have not involved testing.In the following, I will explore a new approach; namely, to transfer the ideals of specificity and testing from the field of linguistics -and science generally -to the evaluation of the diagnostic significance of content for the purpose of dating.On closer inspection, Hávamál contains at least four strong, content-based indications of an early date that have not been subjected to testing in previous scholarship.

Content-Based Criterion 1: Burnt = Buried
The first content-based criterion is found in stanza 81.Hávamál 81.1-2 reads: At kveldi skal dag leyfa | konu er brend er […] ('one should praise the day in the evening [and] a woman when she has been burnt [= received funeral rites] […]').This matter-of-fact statement about praising things only when they are over presupposes that the burning of the dead is sufficiently common for the practice to serve as a recognisable metaphor for death, just as the modern "until she rests in the ground/is dead and buried" draws on inhumation as the typical way of disposing of the in the new skaldic edition (SkP 2: 571-72).Interestingly, Kuhn singled out this occurrence as the only linguistically untenable one, and Kari Ellen Gade's subsequent evaluation in SkP is a good example of how linguistics and textual criticism can inform each other.7.In 21.6 (not listed by Kuhn), the manuscript reads máls, an error through case attraction that is paralleled in stanza 102.9 ok hafða ek þess vaetkis > vaetki vífs.Hugo Gering also lists an instance in stanza 75, but this is based on a daring emendation (Gering 1903: s.v. [col. 760]; Gering and Sijmons 1927-31, 1: 113;von See at al. 1997von See at al. -2019, 1: 679-82), 1: 679-82).
dead.The same metaphor is used in stanza 71.4-5 blindr er betri | en brendr sé ('it is better to be blind than to be burnt [= dead]').The metaphor must presumably have been created in the pagan period, since burning of the dead was a strong taboo throughout the pre-modern Christian period. 8This dating criterion may be tested in various ways.First, it may be compared to the archaeological record, which shows that cremation gradually declined during the tenth century and that it came to a full stop at the end of that century (Taylor 2005: 12).This is promising but not decisive, since metaphors may at times survive their causes for hundreds of years.The abundance of deaths in Old Norse literature provides good preconditions for testing the metaphor's survival.I have explored this by surveying the occurrences of the verb brenna ('to burn').Lexicon Poeticum covers all poetic occurrences, and the digital Dictionary of Old Norse Prose (ONP) and Saga Corpus offer good possibilities for surveying prose occurrences.A search indicates that the metaphor did not survive, since the only superficially similar occurrence is a brend kona ('burnt woman') in the Icelandic Annals of 1301.This is not a reference to burial practices, however, but to her manner of death: Brend kona ór Þýðersku á Norðnesi, ok hálshǫggvin bóandi hennar ('A woman from Germany [was] burnt on Nordnes, and her husband was decapitated') (Islandske Annaler indtil 1578: 52).The woman in question was the so-called False Margrete, executed on Nordnes in Bergen (Mitchell 2022).
As burning now served as punishment and the practice of burning one's enemies in their houses remained the most efficient way of disposing of them, whereas cremation of the dead was abandoned, the old metaphor would gradually have become misleading.The new taxonomy of human incineration is well illustrated by Atlamál, whose ratio of expletive of suggests a date after c. 1000, most likely in the twelfth century, and scholars generally agree that the poem is late.In stanza 39, the deceptive Vingi enumerates his options for killing his opponents: burning (in the house), cutting them down or hanging them (Kommentar zu den 8.The importance of the Christian taboo in a conversion setting has now received additional confirmation through the regulations against cremation in a penitential most likely composed for Willibrord's mission to the Frisians.Rob Meens' brilliant analysis of the text gives us a valuable and more credible testimony to missionary activity among the pagans than, e.g., the texts connected to Saint Boniface (Meens, forthcoming).
Liedern der Edda, 7: 513-14).Stanza 87, by contrast, mentions cremation and specifies this: Brend mundu á báli ('you will be burnt on a pyre').Interestingly, we here see the Christian perception of cremation as a punishment: Brend mundu á báli | ok barið grjóti áðr ('you will be burnt on a pyre and stoned in advance').Atli here predicts that Guðrún will receive the death of a witch, with its concomitant destruction of her remains. 9These negative connotations of cremation are absent from likely early poems, such as Hávamál and Guðrúnarhvǫt. 10 Later authors knew that the pagans practiced cremation, but the crucial point in the present context is that when referring to this practice, they always disambiguated: Þá skyldi brenna alla dauða menn ('then all the dead should be burnt'); Eptir þat lagði Brynhildr sik sverði ok var hon brend með Sigurði ('after that Brynhildr ran herself through with a sword and was burnt with Sigurðr') (Heimskringla 1: 4; Snorri Sturluson,Edda. Skáldskaparmál: 48).Evidently, they felt that this was necessary even when describing the remote past, where references to cremation might be expected.By contrast, Hávamál describes a timeless present, where dying and burning can be used as poetic synonyms without further qualification.If stanzas 71 and 81 were composed by a Christian poet, one would expect the present tense of Hávamál to make the need for disambiguation even greater than in descriptions of the pagan past.The passage in the Icelandic Annals supports this interpretation.Ambiguity would defeat the purpose of stanzas 71 and 81, which require the meaning 'dead' to make sense in the gnomic context.
The 'burnt = dead' criterion has been little used by eddic scholars.In their commentary on the eddic poems, Hugo Gering and Barend Sijmons used it to date stanza 81 to before c. 850, based on now outdated perceptions of the chronology of cremation and without further evaluation (Gering and Sijmons 1927-31, 1: 117) It should be noted, however, that this may not be a useful dating criterion, since largely correct perceptions of early cremation clearly lived on.Rather, it is the rare, anachronistic perception found in Atlamál that may serve as a dating criterion, as well as the metaphor discussed above.
The Dating of Hávamál as an archaic feature (Evans 1990-93: 415).Subsequent major commented editions do not evaluate or avail themselves of it (Kommentar zu den Liedern der Edda; The Poetic 'Edda'; Eddukvaeði).As I hope to have shown, however, the criterion is susceptible to testing along several parameters.

Content-Based Criterion 2: Deyr fé | deyja fraendr
The second content-based criterion is a quotation.Evans 1986: 111-12).Like the couplet in Hávamál, this long line has double alliteration (d : f in Hávamál; f : l in The Wanderer).The Wanderer is an elegy on the transitory nature of earthly glory, and as such, it is comparable to Hávamál stanzas 76-77, although in The Wanderer, the contrast is with the eternal Christian afterlife, not with one's reputation after death.The two poems suggest the possibility that feoh-frēond, perhaps associated with double alliteration, was a formula spread in the North and West Germanic area, connected to the transience of life.This observation might suggest that the couplets in Hávamál and Hákonarmál derive independently from oral tradition, but metrical observations would indicate otherwise.We are not simply dealing with the pair fé-fraendr, but with an identical couplet, and one in which both verses are hypometrical.Most importantly, the first verse contains only two syllables.This feature is found only in ljóðaháttr, and only in the first verse, where it is a rare but apparently permissive licence, with 21 instances in all (Suzuki 2014: 642).The couplet is thus highly marked from a metrical perspective, and it defies probability that two poets would independently make such a marked metrical choice when drawing on a shared tradition.Rather, we are most likely dealing with a loan between the two poems, and this conforms with the overall tendency of skalds to pick up on metrically marked features (e.g. the brestr erfiði Austra-type and type Eε; see Patria 2023).
The next problem then becomes that of deciding the direction of influence.The most important feature for the evaluation of this question may be the fact that Hákonarmál is a narrative poem, its contents being specific to the situations it describes.Deyr fé | deyja fraendr is the only statement in the poem that could also serve as a general comment on the human condition.By contrast, the first and longest part of Hávamál, where this statement is found, is all about the human condition.Allusions or quotations often result in stylistic or semantic inconsistencies, due to the interference of another text. 11The assumption that Hákonarmál has borrowed from Hávamál thus has considerable explanatory power, since it would explain the "gnomic moment" in Hákonarmál, as well as why the borrowing has taken place; namely, to provide the added rhetorical power of an allusion to the universal conditions of humankind, thus portraying the death of Hákon as an event of universal importance.By contrast, it would be difficult to explain why the Hávamál poet -or the poet of the gnomic section -would turn to a poem of such specific and limited scope as Hákonarmál in order to convey a universal message.We see a similar dynamic at play in the allusion from Arnórr jarlaskáld's Þorfinnsdrápa 24 to Vǫluspá 57, where the direction of borrowing can be decided on metrical grounds.The first couplet of Vǫluspá 57 reads: Sól tér sortna, sígr fold í mar (Edda.Die Lieder des Codex Regius: 13) The sun becomes black, the land sinks into the sea.
11. Cf. that Gísli's use of tún to denote a mound would be difficult to explain without intertextual interference (Olsen 1928: 7).

The Dating of Hávamál
A detailed evaluation of the direction of borrowing, including a discussion of variants, is in preparation, but the crucial metrical point is evident in the above quotation: sígr and søkkr are metrically and semantically equivalent, and the only compelling reason to change the verb is that dróttkvaett requires internal rhyme (søkkr : døkkvan) (Males and Myrvoll, forthcoming).By contrast, if the direction of borrowing went from Þorfinnsdrápa to Vǫluspá, there would have been no reason to change the verb.This observation allows us to see a dynamic where a universal feature of the human condition, namely the end of the world, is repurposed to apply to the death of the jarl of Orkney.The assumption that deyr fé | deyja fraendr has been appropriated from a universal to a specific context thus has the support of an attested analogue.
In addition to these observations, it is also noteworthy that Hákonarmál uses the metre of Hávamál -ljóðaháttr -when treating Óðinn and topics related to him, but otherwise uses fornyrðislag (a single exception is found in stanza 2) (SkP 1: 176).This differs from the unclear distribution of the two metres in Eiríksmál, and it suggests that Odinic topics are treated in the metre of Óðinn's own poem, that is, Hávamál.This amounts to a metrical allusion, strengthening the hypothesis of the textual allusion and a date before 961.To the best of my knowledge, none of the three parameters above -conspicuous metrical marking, the overall theme of the poems and the "Odinic" distribution of metres in Hákonarmál -have been evaluated in previous scholarship.

Content-Based Criterion 3: Gjalt(r)
In Hávamál 129 we read that upp líta | skalattu í orrostu | gjalti glíkir | verða gumna synir | síðr þitt um heilli halir 12 ('you should not look up in battle -the sons of men become like a gjalt -[in order that] men be less likely to curse you [þitt = you and things related to you]').
The word gjalt is known from later saga tradition, where the expression is invariably verða at gjalti ('to become mad with fear in armed con-12.Edda.Die Lieder des Codex Regius: 38.Einar Ól.Sveinsson 1952: 168-70, addresses the "problem" of the double couplet before the long line and proposes that an additional long line has fallen out.In light of the shifts between fornyrðislag and ljóðaháttr in surrounding stanzas, however, I am not inclined to see this irregularity as problematic. flict'; Einar Ól.Sveinsson 1952: 171-72).Verða at + dative means 'to become', and originally, the phrase presumably meant 'to become gjalt = to become mad with fear'.Syntactical analysis shows, however, that by the time of our written sources, this is no longer the case, and the absence of such analysis is the most glaring omission in the scholarship on gjalt, both in the sagas and in Hávamál.In the sagas, at no longer serves as a preposition, but rather, at gjalti has become a petrified adverb, since there is no plural *þeir urðu at gjǫltum.Rather, at gjalti is used also of plural subjects. 13If the constituent parts were analysed at all, gjalt must therefore have been taken as an abstract 'madness', and a loose translation 'he/they entered into madness' may give a rough sense of the dynamics involved.Apparently, the underlying semantics had been lost, but this need not be the case in Hávamál.
Only Hávamál contains the detail of looking up in battle so that a curse can take hold, which allows us to see a connection to Irish stories about Suibne geilt (Suibne the madman), who looked up in battle so that the curse of Rónán could take hold (Einar Ól. Sveinsson 1952: 175).
The assumption that gjalt derives from Old/Middle Irish geilt poses a phonological problem that has not previously been addressed.In geilt, both the /g/ and the /l/ are palatal, and in such an environment, one would expect Irish /e/ to be rendered by /e/ in Old Norse.Irish /e/ was rendered by /ja/ in another environment, however, namely when the first consonant was palatal, the second "neutral", as in Kjarvallr, Kjallakr, Myrkjartan, Kerþjallaðr, ingjan and bjannak (the <j> indicates that the preceding consonant was palatal in Irish, and the bold letters are neutral in Irish).For this reason, a form gelt with neutral /l/, like Modern Irish gealt, must have been the starting point for the Norse adaptation. 14There is some debate as to whether the neutral variant is primary or secondary 13.The lexicographer Johan Fritzner proposed that when knowledge of the Irish background was lost, gjalt was reanalysed as a deviant form of gǫltr ('boar, male pig').This does not explain, however, why the singular would be retained for plural subjects.
In addition, Old Norse does not feature variation between /ja/ and /ǫ/, and the supposed analogy in the word svíngalinn ('swine-mad') is hardly valid, since it is found only in Bárðar saga, where we also find svíndrukkinn ('swine-drunk').This suggests that svín-is there used as an intensifying prefix, as in later Scandinavian languages, and that the word svíngalinn has little bearing on perceptions relating to the mental faculties of swine.See Sayers 1994: 175. 14.This form may be secondary, possibly affected by the following back vowel in geltach 'mad' and geltacht 'madness'.
in Irish, but it appears to be attested in the ninth century, and Hávamál indicates that it had developed no later than the tenth. 15There is thus no phonological problem if one assumes that gjalt was based on gelt rather than geilt, and nothing appears to stand in the way of this assumption.The detailed story of Suibne is known from the twelfth-thirteenthcentury text Buile Ṡuibne, which is too late to have influenced Hávamál. 16 Much earlier, however, the so-called Reichenau Primer, dating to in or around the 840s, contains a poem entitled Barr edin ('The Ivied Summit') and attributed to Suibne geilt. 17The poem praises life in the wild -that of Suibne and by extension that of the hermit -and refers to Suibne's/the hermit's abode as maigen 'na áigder rindi ('a field in which spear-points are not feared'). 18The combination of the name Suibne geilt, life in the wild and fear of the battlefield shows that essential elements of the story were present in this early period, and this is confirmed by the poem entered immediately afterwards, which is attributed to Moling and praises the bliss of those who do the will of God (Thesaurus Palaeohibernicus, 2: 294).Moling appears as Suibne's spiritual saviour towards the end of Buile Ṡuibne, and the sequence of poems in the Reichenau Primer thus aligns with that text, albeit with a more pronounced spiritual 15.The earliest evidence shows spellings with and without <i>, but John Carey notes that the forms with <i> are diagnostic (Carey 1984: 95 n. 15).If geilt is a loan from Welsh gwyllt 'insane, wild; madman', which seems likely, the front vowel of that word also supports this (Carey 1984: 95).It may be questionable, however, whether only forms with <i> are diagnostic, except in the very earliest period.In his coming edition of Acaldam Ḟind ⁊ Oiséni (The conversation of Find and Oiséne), David Stifter identifies the Suibne motif, and it seems likely that the exemplar, probably dating to the ninth century, had gelt (the Royal Irish Academy's coming volume on MS 23 N 10).The relevant passage reads: Nícon ralus ó rígluch | gelt for fedaib i ndíthrub […], which translates approximately as 'I have never come across a veteran (= aged warrior) [like a] madman [up] in the trees in the wilderness […]'.16.This is the date usually given to the text, based on its association with The Battle of Mag Rath, which was probably composed before 1197.Linguistically, Buile Ṡuibne could belong anywhere in the period c. 1200-1500 ('Buile Suiḃne': xvi).Nonetheless, some of the poetry in it was likely composed in the Old Irish period (Ní Dhonnchadha 2014: 23-25).17.The manuscript was brought to Reichenau by Martin Gerbert in the eighteenth century and was produced in West Francia, probably in the area of Laon-Soissons. See Bischoff 1981[1977]), 3: 47-50; Tristram 1999.18. Thesaurus Palaeohibernicus, 2: 294.A similar statement -nidom nía (OI nita nía 'I am no warrior') -is found in a quatrain in Buile Ṡuibne that for reasons of linguistic form and independent attestation is likely to go back to the Late Old or Early Middle Irish period (Ní Dhonnchadha 2014: 23-25).
focus.There are thus no chronological impediments to the assumption that the correspondences between Buile Ṡuibne and Hávamál 129 are due to influence from the Irish tradition.Previous treatments of the Old Norse word gjalt have not clearly distinguished three separate contexts: the phrase verða at gjalti, Hávamál 129 and Konungs skuggsjá.In order to produce a clear analysis, the semantics within each context and their relation to each other and to Irish tradition must be evaluated separately.The phrase verða at gjalti is treated above, and I now turn to Hávamál 129.
Among Old Norse texts, only Hávamál refers to becoming like a gjalt, as opposed to simply becoming gjalt.In Hávamál, gjalt is apparently a concrete noun, since only these can serve as points of comparison with other concrete nouns: one can become like a madman, but not like madness.This is unlike prose texts where, if it was analysed at all, gjalt must have been taken as an abstract noun.In addition, the comparison between the plural gumna synir and the singular gjalt in Hávamál suggests that the Irish epithet is treated as a name, and not as a common noun meaning 'madman'.If gjalt were a common noun here, one would have expected *gjǫltum glíkir ('[they become] like madmen'), not gjalti glíkir ('[they become] like a madman'). 19Epithets are often used as appellatives in Norse, so it is natural that Suibne's famous epithet would be treated as his name.Thus, for instance, Eyrbyggja saga chapter 12 explains that Snorri goði's real name was Þorgrímr and that Víga-Styrr's was Arngrímr, and the epithets have otherwise replaced their names, in Eyrbyggja saga and elsewhere.In poetry, we find, for instance, Lúfa for Haraldr lúfa in Haraldskvaeði 10 (c.900) and Skreyja for Eyvindr skreyja in Eyvindr's lausavísa 4 (c.961) (SkP 1: 104; 219 [cf.Eyvindr skreyja in the preceding stanza]).Nothing thus stands in the way of respecting the syntax and reading Gjalti glíkir ('like Gjaltr', with the necessary addition of nom.m. -r = Suibne gelt).We may thus add another exact correspondence to the Irish stories; namely, the name of the protagonist.This strongly suggests that the Irish narrative background was known at the time of the composition of Hávamál.In the phrase verða at gjalti, by contrast, indications of such knowledge are not only absent, but the linguistic usage is incompatible with the Irish narratives.
In Konungs skuggsjá (Norwegian, mid-thirteenth century), a detailed description of the Irish narrative background resurfaces as one of the wonders of Ireland (Einar Ól. Sveinsson 1952: 173-74).We here read of the men called gelt -apparently so called by the Irish -who verða at giellti, are struck by fear in battle and take to the woods, spending twenty years there and growing feathers (Konungs skuggsjá: 61).Not only does this description contain a wealth of information that is otherwise unattested in Old Norse, but Hávamál's two key elements of looking up and a curse are absent.Like many aspects of the account of Ireland in Konungs skuggsjá, this description must therefore be due to new Irish input (on contemporary Irish input in Konungs skuggsjá, see Meyer 1910;Etchingham et al. 2019: 43-121). 20This is also suggested by the spellings gelt and giellti, which indicate a new phonological analysis, presumably based on the form geilt rather than gelt with a neutral /l/, as in Hávamál. 21We are here in the remarkable position of being able to compare two separate imports of the same Irish narrative into Old Norse tradition.For the present purposes, however, it is sufficient to note that the description in Konungs skuggsjá is not indebted to internal Old Norse transmission.It does not, therefore, have any bearing on the assumption that knowledge of the Irish narrative was lost between the composition of Hávamál 129 and the formation of the adverb at gjalti.
In the martially oriented Old Norse tradition, an economical expression for the semantic field of going mad with fear in battle was most li-20.While Meyer considered the input to be exclusively oral, Etchingham et al. think that it is predominantly written.Some of their arguments may be compelling, such as; Diermicius (p.74); Themer (for expected T(h)ever; p. 78); detached genitive singular i-stem (p.81; similarly blaðma pp.68-69).Other arguments are more difficult to accept, e.g. that loch ought to have given luch in auditive approximation (pp.68, 71; luch means 'mouse' in Irish) or that the rendering of Irish -ch-by ON -g-might be a written feature.The argument regarding ec > ee (pp.67-68) does not appear to be valid, since variants with c, as well as confusion between c and t, are attested in the ON transmission.Thus, Logheehag (p.51): logechag (e), logecag (a), logetagh (f); see Engesland, forthcoming.21.The main, Norwegian manuscript (AM 243 bα) has the spellings gelt and giellte.The Icelandic AM 243 e fol.and AM 243 n fol.have geltir and giallti, but in giallti, the a is due to correction of e, and geltir is apparently due to grammatical trivialisation (adding an expected masculine plural ending).The evidence thus unambiguously points to the spellings gelt and giellte in the archetype.See Konungs skuggsjá: 61.
kely attractive.This probably accounts for the detachment of Gjaltr from its narrative background.The process requires at least two stages.The first is one in which (older?) speakers use some economical expression referring to Gjaltr but are understood by other (younger?)speakers to refer to an abstract concept.It is not likely that the expression at gjalti would behave as an adverb at this stage, since linguistic practice would have been affected by speakers who knew that, in fact, Gjaltr was a person.It is only when the second group of speakers becomes the transmitter of linguistic conventions to another generation that the full transformation into an adverb can take place.The setting for this transmission was one of three-generation households, so the first, second and third group of speakers may have had some 50-60 years between them.In addition, narratives are not like phonological changes, that are irrecoverable soon after they have gone through a linguistic community.Rather, at any given time, some people will know a narrative, others will not, and this situation may continue for a considerable period.We may therefore assume that linguistic practices were divergent for a time.For these reasons, it is difficult to believe that the full adverbialisation of at gjalti could have been carried through in less than a hundred years or thereabouts.It is therefore unlikely that a poet could have expected his listeners to be acquainted with key features of the story of Gjaltr in the twelfth century, but that the many texts using the expression verða at gjalti in the thirteenth would only know at gjalti as an adverb, possibly drawing on an abstract noun.If we suppose that the expression verða at gjalti had developed by 1200 at the latest, preceding observations make it likely that stanza 129 was composed before c. 1100.This places us within the period of a pronounced mythological slump, when it is unlikely that any part of a poem with Óðinn as a speaker was composed.This slump began c. 995 or c. 1015 at the latest, and stanza 129 was thus most likely composed before these dates (Males 2020: 39-94).A similar terminus ante quem is suggested by the loan itself, since other poetic occurrences of Irish loans are found in poetry before c. 1000, with the exception of King Magnús berfoettr's use of the word ingjan (= Ir. ingen, 'young girl') to add local colour to his description of his love affair with an Irish woman in Dublin in 1098.Other poetic loans, by contrast, suggest intimate knowledge of matters Irish, and not, like Magnús' stanza, that Irish is something exotic.A number of factors thus combine to suggest that stanza 129 was composed before c. 1000.
While Einar Ól.Sveinsson's 1952 study of Hávamál 129 places a premium on specificity as a precondition for diagnostic significance, Einar left several avenues for testing the stanza's dating implications unexplored.Most notably, the lack of syntactical analysis of Old Norse occurrences of gjalt/Gjaltr has meant that their internal differences have remained hidden.This is the most fundamental precondition for using Gjaltr in stanza 129 as a dating criterion.In addition, Einar did not discuss Irish sources earlier than Buile Ṡuibne, which strengthen the case for early influence.
It is better not to ask than to sacrifice too much; a gift always looks to [= demands] its recompense.It is better not to have sent [a request] than to slaughter too much.
This stanza provides the answer to a question posed in the preceding one: Do you know how to sacrifice?Apparently, the key lies in moderation, since too great a gift to the gods by means of sacrifice would demand too great a recompense, according to the logic of the exchange of gifts.This message is quite unlike the analogues from Hugsvinnsmál and Disticha Catonis adduced in Kommentar zu den Liedern der Edda,1: 909 (cf. SkP,7: 431,.In those poems, sacrifice is to be avoided altogether.22 Whether of a pagan or Christian background, the couplets in 22.The fact that the Kommentar notes these stanzas in Disticha Catonis and Hugsvinns-Disticha Catonis are compatible with a Christian view, whereas Hávamál 145 is not, since it presupposes that one should sacrifice living victims (sóa).There are no injunctions presupposing the sacrifice of living creatures in Old Norse literature from the Christian period, and there are strong reasons for why a Christian would not produce one: Christ's sacrifice has replaced all other sacrifices (Heb 7.27), and the sacrifice of living creatures was associated with Jews and pagans.
The crucial point for using this stanza as a dating criterion is that it is gnomic, encouraging the right course of action.While sacrifices are described in numerous Old Norse texts, the reader is otherwise never, as far as I am aware, encouraged to sacrifice living creatures, whether in moderation or in any other way.Rather, other gnomic or normative texts, such as Hugsvinnsmál or the laws, insist that one should not perform sacrifices to the gods.In medieval and modern historiography alike, authors often describe behaviour which they would not encourage in contemporary society, and gnomic and normative texts therefore speak to the values of a society in a different way than does historiography, whether factual or fictive.Thus, for instance, a saying *"thou shalt kill" would not be a good fit within the Judeo-Christian tradition, despite the many killings described in the Bible, some of them viewed with favour by God.As an injunction, only "thou shalt not kill" is sanctioned by tradition.Similarly, I doubt that immoderation is encouraged in any gnomic tradition, even though immoderate individuals tend to attract the interest of historians.The evaluation of whether gnomic and normative utterances belong to one or the other tradition must therefore be based on comparison with other utterances within the same category, and not with historiography or other narrative.
The injunction to sacrifice within reason has, to the best of my knowledge, not explicitly been used as a dating criterion.Scholars such as Hugo Gering and Finnur Jónsson appear to have taken the pagan composition of Hávamál as self-evident, and additional evidence, such as that provided by stanza 145, would have seemed superfluous to them (see e.g., Sijmons Gering 1927-31, 1: 153;Finnur Jónsson 1920-24, 1: 224-44).By contrast, von See and the other editors of Kommentar zu den Liedern der Edda seem to be so focused on finding similarities in Hugsvinnsmál mál as analogues to Hávamál 145 is one of many examples of their bias for late dating (see Males 2022).
The Dating of Hávamál and Disticha Catonis that they fail to notice that these texts say something very different from Hávamál 145.For various reasons, then, one of the most obvious dating criteria in the poem has largely been left out of the scholarly debate.

Conclusions
As the preceding examples show, content may be subjected to testing no less than formal criteria.Indeed, all four content-based criteria may be stronger than the overall ratio of expletive of, since poets occasionally archaised using this feature.There is thus no absolute correlation between formal criteria and diagnostic reliability.Rather, the most important characteristics of a criterion are its specificity and its susceptibility to testing, the former being a precondition for the latter.
For the long and seemingly composite Hávamál, it is a particular asset that indications of a date before c. 1000 are spread across the poem: stanza 4 (of before noun), 21 (of before noun), 32 (vr-), 38 (of before noun), 71 (brendr sé), 76-77 (deyr fé), 81 (brend kona), 129 (gjalti glíkir), 145 (sacrifice within reason). 23The data therefore invites us to contemplate the possibility that the poem is the product of a single creative mind and that the sections of the poem are intended as such.It is somewhat conspicuous that linguistic archaisms are found only in the first part of the poem, but these features are rare, and there is thus no reason to expect an even spread.As we shall see, there is a factual contradiction between stanzas 13-14 on the one hand and 104 and 109-10 on the other, possibly suggesting that the poet drew on one or two pre-existing poems.The concentration of archaic linguistic features to the first part of the poem may lend some support to such a hypothesis, although coincidence or transmission may be equally plausible explanations of the distribution.It should be noted that vr-and expletive of before nouns appear to have been optional features down to c. 1000, and their distribution may therefore equally well be due to stylistic preferences as to relative chrono-logy.If Hávamál is in some sense composite, I would nonetheless argue that the implications of this background are less prominent than scholars generally assume and that the poem is best viewed as the work of one poet.This calls for some elaboration.
To date, most scholars have viewed the poem as composite, due to its length and divisions into sections (for an overview, see McKinnell 2007).In the spirit of analytical Homeric studies of his day, Karl Müllenhoff first divided the poem into six parts, and scholars have largely followed his lead (Müllenhoff 1891: 250-88).Indeed, David Evans states that it is "inconceivable" that Hávamál was originally composed as one poem (Evans 1986: 7).Indications discussed in this article suggest that the different parts of the poem are roughly contemporary, however.The poem may still be composite, but the analytical value of this claim diminishes if the constituent parts of the poem cannot be told apart chronologically.Furthermore, there is reason to question the assumption that a division into sections need indicate that these existed before the poem in which they are contained.
The lists of dwarves in Vǫluspá and of names of Óðinn in Grímnismál also amount to semi-independent sections, and drápur are divided into bálkar ('sections').There are thus a number of analogues to dividing a poem into segments, and it is doubtful that their presence should be taken to suggest that more than one poet has been at work in the way that, for instance, two linguistic strata would have.Another potential counterindication to unified composition is the fact that the speaker, Óðinn, mainly uses the first person ek 'I', but in stanzas 80, 109, 110, 142 and 145, we find references to Óðinn in the third person, and in stanzas 110, 143, 160, the ek does not refer to Óðinn, but is rather the voice of the poet.Most scholars take these inconsistencies as indications of different strata (von See at al. 1997See at al. -2019, 1: 460-61), 1: 460-61).This conclusion is questionable, however, since a similar situation applies in Vǫluspá, where the vǫlva first speaks in the first person and then shifts to hon 'she', and there appear to be no strong reasons to assume that the "I" and "she" sections represent two different strata.Furthermore, the shift between persons has a literary dimension.In stanza 110, the poet distances himself from Óðinn's immoral behaviour, and in stanzas 80, 142, 143 and 145, we get an "outside" view of Óðinn as the creator of runes.This distribution of functions is not likely to be coincidental, and the assumption that the shifts are simply remnants of older building blocks therefore presupposes a lower opinion of the poem's sophistication than its characteristics suggest.This is most likely attributable to the perception of eddic poetry as "folk" poetry, which guided early Old Norse scholarship and the effect of which remains with us (see the discussion of Müllenhoff below).Indications and counterindications of unified composition and literary sophistication must therefore be evaluated anew, without the unwarranted assumption that eddic poetry is by nature artless and "folksy".
This assumption probably also accounts for the fact that few scholars have drawn attention to indications of Hávamál's unified nature.Thus, for instance, the divine origins of runes/spells are signalled in stanzas 80 and 111 and are then described in stanzas 138-64, and Óðinn's adventures with women in stanzas 96-111 are preceded by a number of statements to the effect that men and women should not trust each other.In stanzas 13-14, Óðinn admits to having been too drunk when he visited Gunnlǫð, and in 104-10, his betrayal of her is described in detail.Stanza 103 presents general advice, as if echoing the gnomic setting of the previous Gunnlǫð stanzas (13-14).Seemingly abrupt changes of topic are in these and other ways foreshadowed in preceding sections, until stanzas 163-64 finally announce that the poem has reached its end.Thus, even if Hávamál may have been compiled from pre-existent sections, the poet must have adapted these so much that it would still make sense to view the poem as the work of one poet.Snorri certainly perceived of Hávamál as a unit when he used it as a framing device for Gylfaginning (Males 2021: 129-32).
This unity is evident also on a conceptual level.Hávamál is a collection of advice interspersed with glimpses into relevant aspects of the life of the divine advisor: his experience with women, which is connected to the topic of tension between men and women, and his acquisition of knowledge, which is what he imparts in the poem.Narrative and advice thus inform each other, and we must therefore assume that the composite character of the poem is intentional.This conforms with what one might expect on general grounds, since the poetic and saga corpus at large suggest that the composition of poetry was seen as an individual endeavour.
With these observations in mind, there remains one indication of the conjoining of two different poems or traditions.In stanzas 13-14, the father of Gunnlǫð is called Fjalarr, whereas in 104 and 109-10, he is named Suttungr.By way of analogy, another giant is named Skrýmir in Lokasenna 62 and Fjalarr in Hárbarðsljóð 26, suggesting that such va-riation may be more likely to occur between poems than within a single one.As we have seen, however, the two Gunnlǫð sections are crafted to mirror each other, so that while the poet may have drawn on two or more earlier poems, his efforts at creating a unitary composition should not be underestimated, and the spread of criteria across the poem allows us to date this creative act to sometime before c. 1000.In spite of the nearconsensus on viewing Hávamál as composite, I have found the internal contradiction of the name of Gunnlǫð's father adduced as evidence only in the commentary by Sijmons and Gering (Sijmons and Gering 1927-31, 1: 87;cf. e.g. von See et al 1997cf. e.g. von See et al -2019, 1: 531;, 1: 531;The Poetic 'Edda', 3: 51;Eddukvaeði, 1: 325;Evans 1985: 81;Sayers 2015: 394, 397).The reason for the absence of this argument among scholars following Müllenhoff's lead is presumably that Müllenhoff considered stanzas 13-14 a later addition and therefore excluded them from analysis.Today, scholars would generally avoid such an invasive approach, but the debate has nonetheless been conditioned by Müllenhoff's choices.
The three enlarged initials in the manuscript are sometimes adduced as evidence of the poem's composite nature (e.g.McKinnell 2007: 75-76, 106).This feature probably has little bearing on the issue, however, since the variants Fjalarr/Suttungr, if anything, suggest a dividing line somewhere between the first and second initial, and since Snorri clearly perceived of the poem as a unit.After some three centuries of oral transmission, scribes presumably had no better clues than we do for reconstructing the lost building blocks of the poem, nor is it very likely that they would attempt to do so, since this kind of "analytical" criticism is a product of the eighteenth and nineteenth centuries.Rather, the scribe(s) presumably felt that the poem fell into thematic sections and signalled this by their use of initials.
These observations lead us back to the possibility of pagan appropriation of aspects of Christ's crucifixion in stanzas 138-45.Since good dating criteria are spread across the poem and there are no specific indications pointing to a date after c. 1000, these stanzas were most likely composed before that date.More specifically, stanza 145 contains the fourth content-based dating criterion, after which follows an analogue to Christ: þar hann upp um reis | er hann aptr kom ('there he rose up when he returned').Hávamál 138-45 presents us with a sacrificed god hanging on a tree (note that the Cross was commonly understood as a tree in the Middles Ages; Lassen 2009), given no food or drink (cf. the mocking offer of gall and vinegar to Christ; Males 2013: 111-12) and finally returning and rising up (note that rísa upp and the noun upprisa are the two words commonly used for Christ 's resurrection;Fritzner 1883-96, 3: 115, 797).Central elements in Christ's and Óðinn's hanging thus overlap, and no such complex of elements appears to be found in neighbouring cultures, which suggests that Hávamál 138-45 have been influenced by Christian perceptions.Nonetheless, the stanzas are so different from descriptions of Christ's crucifixion that attributing them to a Christian poet creates more problems than it solves.Specifically, as we have seen, the last correspondence is preceded by a statement that is incompatible with Christian views on sacrifice.Stanza 145.1-5 may thus serve not only as a dating criterion for Hávamál and for the "hanging" stanzas, but it also suggests that even if stanzas 138-45 may have been influenced by Christian perceptions, it is unlikely that they were composed by a Christian poet.Furthermore, the traditional elements of the "hanging" motif, including the spear and the hanging, should not be underestimated, but rather, it was presumably the presence of such features that made the addition of Christian elements seem viable in the first place (Males 2013: 111). 24 A more precise date for Hávamál than before c. 961 (the date of Hákonarmál) must perhaps remain elusive, but some general observations may be made.850-900 are the rough dates of our earliest preserved skaldic poems, and the lines of transmission of skaldic and eddic poetry would probably have been largely the same from the time of composition until they were committed to writing.Germanic analogues and the Rök stanza (c.800) indicate that the fornyrðislag tradition is at least about two centuries older, but the corpus preserved in later manuscripts was dependent on its carriers.It seems likely that ljóðaháttr -the metre of Hávamál -developed out of a variant of fornyrðislag during the ninth century, although this assumption must remain somewhat tentative 24.One may note the kenning galga farmr 'cargo of the gallows' for Óðinn in Háleygatal 1 (c.986), providing another early witness to Óðinn as himself hanged, rather than as the god of the hanged, but it is possible that Háleygatal is influenced by Hávamál in this regard.Óðinn's access to hidden knowledge is a prominent feature of stanzas 138-45, the traditional character of which is attested in the rest of Hávamál, as well as other eddic poems.Indeed, this seems to be the main reason for Óðinn's hanging, whereas that of Christ's is the redemption of humanity to the possibility of eternal life.The causes of the two hangings are thus fundamentally different.(Males 2023a).I see no reason to assume that carriers of the poetic tradition would have favoured poems that were considerably older than the skaldic poems simply because the eddic tradition itself was older.The crucial factor is continuous transmission such as that provided by the courts and their poets, and it seems likely that these poets would have memorised poems favoured at their courts and, perhaps, produced by the very poets that we know from their skaldic production.The only poet who can be dated to around the middle of the ninth century, Bragi, has many conspicuous archaic features, and we find no such extreme archaisms in Hávamál (Males 2023b).Such features could, of course, have been obliterated in transmission, but overall, a date c. 900-50 seems plausible, although a cautious scholar might wish to expand the timeframe to c. 850-960.
A plausible dating helps us flesh out the picture of tenth-century elite culture, but also to get to grips with the text itself.Much of the advice in Hávamál is staunchly practical: do not get too drunk, be a friend to your friend, be neither stupid nor too wise, etc. Unsurprisingly in such a context, we also get a variety of "the early bird catches the worm", but one that is decidedly odd to modern ears.Hávamál 58 reads: Ár skal rísa sá 25 er annars vill fé eða fjǫr hafa; sjaldan liggjande ulfr laer um getr né sofandi maðr sigr.
He who wishes to take another's wealth or life must rise early; seldom does the wolf get a thigh lying down, nor a sleeping man victory.
The modern Scandinavian saying runs morgonstund ger guld i mund ('the morning hour gives gold in the hand') 26 and the like, replacing killing and victory with more ordinary work ethics.The injunction "thou shalt not kill" is deeply embedded into Christian culture, and while this ideal 25.MS ri | sa er. 26.Today, most Scandinavians would take this to mean 'gold in the mouth (mun/munn)', since mund (hand) is an obsolete word.
does not always correspond to reality, Christian endorsement of violence generally insists on its religious motivation.Stanza 58 thus appears to evince non-Christian ethics.Even so, I have not included it among the dating criteria, since its relatively unspecific contents make it difficult to test its diagnostic significance.Furthermore, it complicates matters that in the warrior ethic of poetry in particular, exhortations to fight bravely are commonplace in pagan and Christian culture alike.This stanza is different, however, addressing the practicalities of one man taking the life of another and stressing not bravery, but vigilance.With a plausible dating of the poem, a likely explanation of its pragmatic approach to killing becomes evident; namely, that the stanza was not intended to conform to Christian ethics.The practicality of this stanza is matched by that of stanza 145, where even the gods are part of a strictly commensurate system of gift giving and should not be forced to reciprocate beyond reason.
One is reminded of Egill's reaction when Einarr skálaglamm gave him a shield (Egils saga, ch.78).Egill assumed that Einarr was trying to force him to compose a poem in recompense for the shield, and he became so outraged by this that he decided to kill him.In the event, Einarr had left in the morning, and so Egill had to settle for composing a poem instead.Lack of reciprocation was not an option.This is obviously a humorous account by the saga author, but it must have been meaningful to its audience, and stanza 145 shows us just how deeply imbedded the concept of reciprocation was in Old Norse culture.With a plausible dating, stanzas like Hávamál 58 and 145 allow us to flesh out the cultural background of some of the traditional notions that later saga authors could draw on.The dating of Hávamál thus informs our analysis not only of perceptions among the pagan elite in the tenth century, but also of the preconditions of later saga literature.Bischoff, Bernard. 1981[1977].Irische Schreiber im Karolingerreich.In:
The verses Deyr fé | deyja fraendr ('cattle dies | relatives die') in Hávamál 76-77 are also found in Hákonarmál 18, composed shortly after Hákon góði's death in 961(SkP 1: 193).The presence of an identical couplet in both poems is not likely to be coincidental, and two logical possibilities present themselves.Either both poems quote from a shared tradition, or one has borrowed from the other.As David Evans notes, the Old English poem The Wan- derer contains the long line (l.108) hēr bið feoh lǣne, hēr bið frēond lǣne ('here possessions are transitory, here friend is transitory') (Hávamál,