Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? Frog* Abstract. This article explores patterns of language use in oral poetry within a variety of semantic formula. Such a formula may vary its surface texture in relation to phonic demands of the metrical environment in which it is realized. Metrically entangled ken- nings in Old Norse dróttkvætt poetry provide material for a series of case studies focus- ing on variation in realizing formulae of this type. Old Norse kennings present a seman- tic formula of a particular type which is valuable as an example owing to the extremes of textural variation that it enables. Focus will be on variation between two broad semantic categories in expressing the formula’s consistent unit of meaning that are otherwise unambiguously distinct: proper names for mythological beings and poetic terms for weapons and armour. This article introduces an approach to kennings as semantic for- mulae and includes an illustrative case study on kennings meaning ‘battle’ in the last three metrical positions of a dróttkvætt line. The case study is simultaneously used to demonstrate the degree of integration of mythological proper names in the poetic reg- ister. This article contains only the first case study of a series. It provides foundations for examining variation in the associative links exhibited by names of mythic beings as a category according to the metrical positions in which a battle-kenning is realized. Key words: oral poetry, variation, formula, skaldic poetry, dróttkvætt, kenning This is the first part of a four-part discussion that considers the vocabulary of proper names associated with mythology as potentially being fully inte- grated into the oral-poetic register (language as used in the poetry) of medieval Icelandic skaldic verse. The overall discussion is an exploration of the possibil- ity that, like other nouns in this register, the lexicon of proper names func- tioned in generative composition1 as a resource for meeting metrical demands * Author’s address: Frog, University of Helsinki Folklore Studies, Department of Philosophy, History, Culture and Art Studies, PL 59, 00014 Helsingin Yliopisto, Finland. E-mail: mister- frogfrog@yahoo.de. 1 The expression ‘generative composition’ is used here to distinguish generative language production within the tradition of poetry from objectively conscious composition (even if, in practice, these are extremes on a spectrum). Generative composition thus highlights reliance on internalized and intuitive understandings of the meter and the poetic register, with its con- structions and conventions for realizing the meter, in contrast to an objectified handling of the register with a more analytical apprehension of the meter. Studia Metrica et Poetica 1.1, 2014, 100–139 doi: dx.doi.org/10.12697/smp.2013.1.1.06 mailto:misterfrogfrog@yahoo.de mailto:misterfrogfrog@yahoo.de http://dx.doi.org/10.12697/smp.2013.1.1.06 101 of alliteration, rhyme, syllabic quantity and syllable-count. Recognizing that the lexicon of mythology is integrated into the register of skaldic poetry is not a surprising finding in itself. However, the use of this lexicon has generally been taken for granted: the meter and the language of the poetry tend to be acknowledged as complementary aspects of the skaldic poetic tradition but receive focused address in isolation from one another rather than focus on their interface.2 Little or no consideration has been given to how this area of skaldic vocabulary functioned in versification or the relationship of metrical circumstances to word-choice in the generative production of new verses.3 This is probably in large part owing to the methodological problem of deter- mining the influence of meter on word-choice (cf. Marold 1983: 43). This problem is relevant to individual cases but it is alleviated when attention is shifted to social patterns of language use. The present discussion examines verbal variation in a metrically entangled type of semantic formula called a kenning in dróttkvætt-meter poetry. Although this discussion is focused on a variety of Old Norse poetry that is exceptional in its metrical complexity and its conservatism in transmission, this poetry’s exceptional qualities make the corpus a valuable testing-ground for investigating certain types of variation that will be of general interest in research on formulaic language. Although the Old Norse kenning is a semantic formula of a particular type, this discus- sion has potential to elucidate the phenomenon discussed here as a semantic formula more generally. In the background of the present discussion is an exploratory pilot study on formulaic language in skaldic dróttkvætt poetry (Frog forthcoming). The pilot study tested the possibility that kennings might become ‘metrically entangled’, although these metrically regular formulae might be concealed beneath a surface of synonymic variation as a function of skaldic diction. 2 E.g. Meissner 1921; de Vries 1934; Fidjestøl 1997; Árnason 1991; see also note 7 below. 3 A relationship between the richness of the skaldic register and the demands of the dróttkvætt meter is generally recognized – e.g. Kari Ellen Gade (1995: 3) observes: “although it is not pos- sible to trace the origin of the kenningar, it is clear that the nominal kenning system is intimately connected with the strict formal requirements of skaldic meter, and that the kenningar provided the skalds with the tools they needed to fulfil the requirements imposed by rhyme, alliteration, syllable counting, and sentence structure.” Nevertheless, this relationship frequently remains implicit (cf. Frank 1985: 163–164) or only noted in passing (cf. Meissner 1921: 36n; Sullivan 2008: 32–33). The practical and functional aspects of language are easily overshadowed by considera- tion of their aesthetics (e.g. Frank 1978: 33–54 and passim; cf. Kuhn 1983: 221–222; Whaley 2005: 487–488; Clunies Ross et al. 2012: lxxiii) or by consideration of how kennings work (e.g. Lindow 1975; Amory 1982; Sverdlov 2003). The question of how meter relates to word-choice has never really penetrated the discourse and generally remains outside of consideration. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 102 Frog Research on formulaic language and compositional practice in oral poetry tends to remain rooted in Oral-Formulaic Theory (also called Oral Theory), building from the early definition of the formula proposed by Milman Parry (1928: 16): “an expression which is regularly used, under the same metrical conditions, to express a particular essential idea” (also Lord 1960: 4). Such formulae can be considered metrically bound insofar as they are exclusive to a particular metrical context and will not ‘naturally’ occur outside of that metrical context or are otherwise specific to and invariable in it. The present approach considers this degree of fixity as an extreme on a spec- trum of metrical entanglement. The term metrical entanglement is here used to describe the phenomenon by which use of the lexicon of an oral- poetic system becomes bound up with metrical positions or other metrical parameters. Items in the poetic lexicon, including semantic formulae, can be considered metrically entangled if there is reason to believe that they are conventionally associated with one or more metrical contexts and are uncommon outside of that metrical context or those contexts, even if it is not necessarily striking when they do so. With the assistance of the Skaldic Database, the pilot study was carried out on a random (from the perspective of the study) dataset of 340 metrically situated kennings with the referent ‘battle’ (cf. Parry’s “essential idea”). The study revealed that kennings exhibit clear metric-structural ‘types’. A metric-structural type was defined in terms of the metrical positions filled by each element forming the kenning. Within the dataset, 10 ‘basic types’4 accounted for more than 70% of the examples. Individual battle-kenning types also exhibited, for example, conventions of lines in which they would occur that could not be accounted for by metrical factors alone.5 This indicated that battle-kennings were not always inserted freely into the meter. The 10 most frequent basic metric-structural types were deemed reasonably well attested and with a sufficient relative frequency to be considered to reflect metrically entangled but verbally variable seman- tic formulae. Furthermore, individual types also exhibited lexical prefer- ences and/or were associated with lexical collocations such as rhyme-pairs (i.e. associated with phonic demands of the meter), only one part of which would be an element in the kenning. These suggested varying degrees of 4 On the distinction between ‘basic types’ and ‘complex types’, see Frog forthcoming. 5 E.g. the most frequent battle-kenning basic type (YX3456, with 60 examples) was specifically associated with even lines, although kennings with other referents of this basic metric-structural type appear in odd lines. 103 ‘crystallization’6 of the lexicon in relation to metrically entangled semantic formulae that could advance toward metrically bound formulae, approach- ing Parry’s early description. Such collocations could appear specific to a basic type rather than distributed across types, making it improbable that these patterns were merely an accident of convergent probabilities in the dataset. In other words, the lexicon could also be metrically entangled with a basic type, including words that were not part of the formula itself. This suggested compositional resources that could allow types of flexibility in realization (most of which do not concern us here). The first three parts of the present discussion are case studies of metri- cally entangled semantic formulae of dróttkvætt-meter poetry. Each of the three case studies will focus on the occurrences of proper names in a dif- ferent metric-structural type of semantic formula meaning ‘battle’ (types 123XYy, 12YyXx and YyX456, respectively). These were three frequent types according to the pilot study. They are here addressed in larger datasets that have been extended laterally through the corpus for the identification of as many examples as possible. The examples surveyed should not, however, be considered exhaustive of the basic type, nor do they include variations on the basic types addressed (although a few cases of expansion into a complex type are introduced for discussion). The first half of the present article intro- duces the approach to the material and to kennings as semantic formulae. The second half presents the opening case study which has been chosen to offer an accessible introduction to the metrical entanglement of kennings in dróttkvætt. The two later case studies will illustrate different ways in which the lexicon of mythological proper names becomes metrically entangled in the realization of a basic type. The fourth part in this series will situate the case studies in relation to one another, discuss certain ramifications of the findings and consider their potential to provide a foothold for future investigations. Discussions of skaldic diction generally address kennings in terms of lexicon, syntax, structure, referentiality, ambiguity, interpretation, aesthetics 6 Anna-Leena Siikala (1990 [1984]) advanced the term ‘crystallization’ for approaching rela- tive fixity in the reproduction of traditional narrative related to the degree of fixity in individual memory, which can then be inferred as characteristic in social patterns reflected in larger cor- pora. The term emphasizes relative fixity rather than suggesting binary opposition between ‘fixed’ and ‘free’ or that these are ideals rather than extremes. Within studies of oral poetries (and especially of Old Norse poetries), this is significant because of long-standing presumptions of a binary opposition between ‘memorization’ and ‘improvisation’ (see Frog 2011a: 23–24, 51–54), whereby skaldic poetry becomes ‘memorized’, ‘fixed’ and suffers ‘corruption’ rather than ‘variation’. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 104 Frog and/or mnemonics, as well as typologies, systems and historical development within those systems.7 The present study differs from these in its focus on pragmatic variation in lexical choices in particular metrical contexts. Rather than addressing how kennings refer, the case studies concern language use in realizing, as practice, formulaic expressions that are associated with a particu- lar referent. Rather than focusing on variation between terms or names within a category of semantic equivalence (cf. Frank 1985: 168–169), the focus here engages questions of variation between categories of semantic equivalence. This is done with emphasis on names associated with mythology. This is a vocabulary that has been regarded as particularly marked in use as well as receiving special attention in scholarship as source material for research on vernacular mythology.8 The discussion opened here will consider the degree to which these names were integrated into the lexicon of the poetic tradition and whether, as a part of pragmatic variation in the practice of realizing semantic formulae, a valkyrie could alternate with a spear. Kennings and heiti A kenning is a rhetorical figure of two nouns in a syntactic relationship indi- cating a third, nominal referent. One noun, called a base-word, is open to morphological variation according to its use. This is complemented by a sec- ond noun, called a determinant, in the genitive case or as an uninflected stem 7 E.g. Meissner 1921; Marold 1983; Lindow 1975; Amory 1982; Frank 1978; Gade 1995; Fidjestøl 1997; Sverdlov 2003; Holland 2005; Sverdlov 2006; Würth 2007; Birgisson 2007; Potts 2011; Osborne 2012. In her study of syntax in dróttkvætt, Kari Ellen Gade (e.g. 1995: 107) observes a few cases of kenning constructions characterized by the metrical distribution of their constituents in a line, but she does not explore the relationship of these constructions to lexical variation although she analyzes them in relation to metrical structures. Some studies have addressed the distribution of elements constitutive of kennings across lines of a stanza (e.g. Kuhn 1983: 224–228; Sverdlov 2003; Wills 2009; n.d. a; n.d. b; cf. Sverdlov 2006), but these do not extend to addressing metrical positions of kenning elements within the line or lines. 8 Names of Odin and valkyries tend to be interpreted in terms of mythological significance or use with reference to particular mythological narratives. Little or no consideration is given to the metrical conditions under which such names appear or to their relationship to the poetic system more generally, unless this is to determine the correct/historical form of a name or whether a proposed name is based on a textual error or misinterpretation. Cf. e.g. Falk 1924; de Vries 1934; 1956–1957; Turville-Petre 1964: 61–63; Price 2002: 100–107, 337–346; Abram 2011; cf. however also Lindow 2001: 250. 105 preceding the base-word to form a compound.9 Thus, when the base-word hríð = ‘storm’ is complemented by the determinant sverðr = ‘sword’, it forms a kenning with the referent ‘battle’. This can take the forms hríð sverðs = ‘storm of the sword’, hríð sverða = ‘storm of swords’ or sverðhríð = ‘sword-storm’. This rhetorical figure can be approached as a type of construction – i.e. a “con- ventionalized pairing of form and function” (Goldberg 2006: 3). Thus when two relevant nouns are encountered in an appropriate syntactic relationship (and context of language use), the recognisability of the construction leads the expression to be interpreted as a kenning. Skaldic kennings are never- theless concentrated within certain subject domains and have limited sets of conventional referents. In addition, base-words and determinants used in ken- nings for a particular referent also tended to be representatives of conventional semantic categories, in which lexical variation was enabled according to “para- digmatic substitution” (Clunies Ross et al. 2012: lxxi). (See also Meissner 1921; Fidjestøl 1997.) Lexical variation was enabled by a rich vocabulary of semanti- cally equivalent poetic terms called heiti = (lit.) ‘that which something is called; name’. In the semantic formula ‘storm of swords’, the base-word could be filled with any of a large number of weather-heiti, such as él = ‘snow-shower’, drífa = ‘a fall of snow, sleet’, hregg = ‘rainstorm’, regn = ‘rain’, skúr = ‘shower’, veðr = ‘weather; wind’, etc. Potential for variation expands considerably because other, additional semantic fields with their own sets of terms can equally fill this 9 As either noun in a kenning can be a more developed Noun Phrase, including adjectives, article or number, or may even be another kenning, the rhetorical figure can be abstractly described as realizing two syntactic constructions: NP2-GEN NP1 or NP2-NP1 = NP3. The con- straint that the referent should not be semantically the same as the base-word or the determinant (i.e. NP3 ≠ NP1 or NP2) is relevant to the consideration of mythic personal names in kenning constructions. A battle-heiti cannot be used to form a battle-kenning, but a personal name that is also a battle-heiti is acceptable (e.g. valkyrie-names like Hildr, Gunnr; cf. also Meissner 1921: 73–74, 201–202). This ostensibly produces ambiguity regarding whether certain words are in fact personal names. However, if they are not names, the construction is not formally a kenning. This does not mean that inconsistent uses and interpretations did not occur (cf. styrr as a base-word in Anon Líkn 34VII.6, 8). However, the social pattern of use of these lexemes in battle-kennings can be considered historically rooted in their interpretation as personal names for formal reasons. Although individuals could flex their interpretations of those terms in rela- tion to ideology (cf. Osborne 2013), formal continuities in patterns of use within the register can be reasonably considered to reflect earlier uses of the personal name. Conventional formulaic expressions, such as those explored here, may be generatively produced and/or interpreted without a need to resolve whether a conventional lexeme is a battle-heiti or a personal name (Wray 2002: 130–132; cf. Lindow 1985: 27–28). The present investigation is on the realization of conventional formulae through the register, and therefore the ostensible ambiguity of individual cases is eclipsed where these appear to be historically rooted in conventions of social practice. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 106 Frog function, producing the kenning as ‘noise of swords’, ‘voice/song of swords’, ‘meeting of swords’, ‘game of swords’, etc. The determinant could similarly be filled with other sword-heiti, such as hjǫrr = (poet.) ‘sword’ or even a recog- nizable proper name of a specific sword, such as Laufi, the name of the sword of the hero Bǫðvarr bjarki. Battle-kenning determinants such as egg = ‘edge, blade’ and malmr = ‘metal’ easily blur into a broader category of ‘weapons’ and are not necessarily specific to ‘sword’. Other determinants such as hjalmr = ‘helmet’ and brynja = ‘armour’ can be considered in a broader category of ‘armour’, which can in turn be seen with ‘weapons’ as a still broader category of ‘implements of battle’. Personal names of valkyries, Odin-heiti, names of mythic heroes and kings can be similarly viewed in equivalence classes advancing to a broad category of ‘mythic agents of battle’. This category nevertheless remains semantically distinct from ‘implements of battle’, even if these potentially have functional equivalence in realizing battle-kennings. In any case, the recognis- ability of the construction enables unfamiliar kennings to be resolved when encountered in a contextually appropriate environment, such as bekkþiðurr = ‘stream-partridge’ = duck (Egill Lv 2V (Eg 5).5)10 or viðbjǫrn veggja11 = ‘wood- bear of the wall’ = mouse (Anon Bjark 7III.3–4), as well as kennings for more common referents that may deviate from conventional patterns of paradigmatic substitution, such as hǫlða morð = ‘murder of men’ = battle (Eskál Vell 20I.3). A Usage-Based Approach to Language and Variation The present investigation is developed on a usage-based approach to language and variation. Register is here used as a term and concept for approaching contextually-based variation in language. The term ‘register’ developed in linguistics to designate language as it is used and realized in accordance with a communicative context, thus a register is one conditioned variation 10 Citations are by sigla and stanza numbering of the Skaldic Database. Citations are with reference to the published editions of the associated Skaldic Poetry of the Scandinavian Middle Ages edition where these were available at the time of writing (Clunies Ross 2007; Gade 2009) and with reference to Finnur Jónsson’s critical edition (1967) where they are not. 11 It may be noted that this mouse-kenning may not have been unique to the verse in which it is preserved. Richard Perkins (1982–1985: 169–170) points out that in 19th century folklore, veggja dýr = ‘animal of the walls’ was used as a circumlocution for ‘mouse’ when at sea. This was part of an avoidance vocabulary owing to taboos against using certain words because certain words (like ‘mouse’) could have negative effects or consequences. Thus, this kenning (or associ- ated semantic formula) could have been conventional outside of verse. 107 of language among others (see e.g. Halliday 1978; adapted to the analysis of oral poetries centrally by Foley 1995; 2002). Register will here be used as a tool for approaching variation in how lexical items are handled and inter- preted in relation to poetic systems and poetic genres as socially established contexts of variation in language use. In oral poetries, meter only exists as an abstraction realized through the rhythmic patterns exercised in elocution, maintained through a conventional mode or modes of expression in social practice. The mode of expression consequently shapes the register – the language used to realize those rhythmic structures – as a historical process.12 This has long been acknowledged in uses of formulaic expressions (see e.g. Foley 1991) and is also recognized as a significant factor in the development of the vocabulary of semantic equivalence classes for meeting compositional constraints (cf. Roper 2012). The constraints imposed by the system narrow and focus general linguistic processes so that, to borrow the pseudo-maxim of John Miles Foley (2002: 127–128): Oral poetry works like language, only more so. Accordingly, the register of an oral poetry simultaneously reflects outcomes of linguistic processes and has the potential to offer insights into those areas. The present investigation is of the register of dróttkvætt, the meter that is considered characteristic or even iconic of so-called skaldic poetry, in which it is associated with a number of genres. Old Norse scholarship cus- tomarily distinguishes between ‘eddic’ (folk) and ‘skaldic’ (court) poetries, but this binary distinction is an artificial construct that is more intuitive and practical than scientific; ‘eddic’ and ‘skaldic’ cannot be considered for- mally defined, analytical categories (see e.g. Frank 1985: 159–160; Clunies Ross 2005: 21–28; Frog 2009: 229–231). The Old Norse poetic system was extremely dynamic and is better approached as systems of registers, modes of expression (through which meters were realized) and prosodies (includ- ing syntactic complexity) fulfilling conventional constellations according to 12 Halliday did not address oral poetries. In his adaptation of the concept, Foley treated register as generally inclusive: he saw meter as part of register without engaging the concept of ‘mode’. However, if register is approached as a system of signifiers that can be adapted and manipulated across modes of expression, it may be more efficient to acknowledge meter as an aspect of the mode of expression through which the register is realized (see also Frog 2012: 52–54). The dependence of meter on a mode of expression for long-term continuity in an oral tradition is highlighted by the fact that when one mode of expression is exchanged for another (e.g. in dictation rather than a conventional performative mode), registral expressions may be main- tained but the meter will be impacted, whether loosening, tightening or dissolving altogether (cf. Salminen 1934: 200–203; Lord 1991: 41, 45–49; Honko 1998: 81–88; Saarinen 2013: 40). Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 108 Frog genre and application.13 Register is a flexible analytical tool, and its sensitivity and scope can be calibrated in relation to the research object. It is therefore possible to address a common register of Germanic epic poetics,14 a broad Old Norse poetic register shared across eddic and skaldic verse,15 or to distinguish registers of skaldic poetics and of eddic narrative genres, as well as still more sensitive distinctions of particular genres or meters. Conventions of language use in Old Norse poetry varied in relation to meter (Wills 2009).16 The role of register in realizing metrically well-formed expressions interfaces formu- laic expressions with meters. Consequently, metrically entangled formulae are characteristic of particular meters and do not move freely between them, even if they are metrically viable across those contexts (cf. Frog 2012: 52–54). At the same time, characteristics such as kenning density and complexity of syntax could vary considerably across genres and uses of a single meter. This 13 This differentiates the group of Old Norse poetries under the umbrellas of ‘eddic’ and ‘skaldic’ from many oral-poetry traditions to which register has been applied, where a meter is consistent for modes of expression and different genres, where there is a co-variation between genre and meter, or where a genre or meter is in focus and alternative resources do not require address (e.g. Foley 1995; Harvilahti 2003; Tarkka 2005; Sykäri 2011; cf. Stepanova 2012; Frog 2013). 14 This sort of broad perspective on diverse regional and cultural reflexes of a common oral- poetic register has been addressed by Eila Stepanova (2012: 264; forthcoming) as the ‘general register’ of a widespread tradition when addressing the genre and poetics of laments in Finnic cultures. 15 This would be narrower than Lauri Harvilahti’s (2003: 90–115) ‘ethnocultural substrate’, which accounts for the constitutive resources of the full breadth of oral-poetic traditions in a cultural region, or what Stepanova (2012: 281; forthcoming) addresses as ‘supra-registral’ ele- ments and features. The resources described by an ethnocultural substrate form conventional constellations which are socially recognizable as belonging together – i.e. as registers of a par- ticular poetry genre, meter or group of genres – while supra-registral elements and features are shared across these constellations but do not themselves form a constellation characterizable as a register. The register of (attested) eddic and skaldic poetries is representative of only certain parts of the poetry traditions of the culture associated especially with men and bound up with a heroic ideology. Women’s genres of poetry are not preserved in documented texts (although some skaldic verses by women are preserved). It is nevertheless probable that there was, for example, a vernacular tradition of lament poetry sharing features with other lament traditions linked to the Circum-Baltic contact zone (Stepanova 2011: 140), and which may have been distinct from the poetic systems of male-dominated genres (cf. Frog, Stepanova 2011; Stepanova 2012). 16 Ragnar Ingi Aðalsteinsson (2005) has observed that alliteration fell on different parts of speech in dróttkvætt and the eddic fornyrðislag meter. However, the examples of alliteration in (early) fornyrðislag were drawn from eddic narrative genres and did not extend to use of the same meter by skaldic poets. It therefore remains ambiguous whether this variation is a func- tion of meter or genre (cf. corresponding variation in genres of Kalevala-meter poetry: Leino 1970: 132–137, 186). 109 is significant when the same meter could be characteristic of several genres and uses.17 Thus, skaldic poets could also employ ‘eddic’ formulae when using meters associated with ‘eddic’ genres.18 At the same time, a formulaic expression characteristic of a particular meter could maintain formal continuity across genres while varying in semantics according to the narrower generic regis- ter (Frog 2013). In this tradition, registers of genre and meter could intersect and diverge: a single composition could employ multiple meters and the same meter could be used across multiple genres as well as having extra-generic uses (Frog 2009). When turning attention to the register of a particular meter, this means focusing attention on an area within a broader register associated with a number of genres, which could also allow it to be described as a ‘sub-register’ or a special type of “context-dependent lexicon” (Stepanova 2012: 264–265) within the broader register. This approach appears valid for Old Norse poetries because the corpus is characterized by registral resources conventional across genres while being characteristic of certain meters and not others. The area that is being distinguished is characterized by the mode or modes of expression interfaced with the particular meter as a determinant condition on register use (cf. Halliday 1978: 61–64).19 Those mode(s) reciprocally shaped those resources while the strategies for realizing metrically well-formed expressions develop historically in relation to social use across different genres. Kennings as Semantic Formulae It was proposed above that kennings could be approached as a type of construc- tion, through which form conveys information about the function or meaning 17 There are also potential indicators that skaldic uses of ‘eddic’ meters employed a different mode of expression. This can be inferred from indications that many skaldic uses of these meters (ideally) realized them with syllable-based rhythms rather than stress-based rhythms (cf. Turville-Petre 1976: xvii). 18 Cf. Thorvaldsen 2008: 154; but on differences in syntactic structures, see also Mellor 1999 [2008]: 156–160. 19 Within an oral-poetic tradition, the characterization of such a register by formal constraints allows it to be approached centrally through the system of signifiers as resources in expression rather than what these signify, which may vary considerably according to genre and context of use. The ambiguity of formally conventional expressions and symbols enables flex and adapt- ability that may be fundamental to their ability to remain interesting and compelling resources, especially through significant changes in social and semiotic environments (cf. Bell 1992: ch. 8; Osborne 2013). Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 110 Frog of an expression. Constructions exist on a continuum from abstract grammar and rhetorical figures to lexicalized idioms. Formulae constitute one area along this continuum. Rather than determining whether something ‘is’ or ‘is not’ a formula in Parry’s terms of fixity, the area of formulae along this continuum can be approached in terms of degrees of crystallization that may be quite fluid. A formula is approached here is as a broad linguistic phenomenon, which, in oral poetry, occurs within a metrical context (cf. Kiparsky 1976), rather than defining the poetic formula separately, as was done across the early stages of Oral-Formulaic Theory. Alison Wray (2002) defines the linguistic formula in terms of morpheme- equivalence: it is characterized by a coherent unit of meaning with an exclu- sive entry in the mental lexicon of language users, even if it can be analyzed and appropriately interpreted according to rules of the grammar.20 Parry’s verbal expression communicating an “essential idea” correlates with Wray’s morpheme-equivalence conjoined with metrical boundedness. The emphasis on metrical fixity in defining the oral-poetic formula proved overly narrow even for the poetic traditions on which the definition concentrated (see e.g. Hainsworth 1968; Foley 1991; Harvilahti 1992a). The model was based on long epic forms of South Slavic and Homeric traditions and it could not be effectively applied to other traditions without adaptation.21 The early approach emphasized so-called ‘composition in performance’, focusing on traditions in which the verbal level of the text was realized more or less freely in the pro- gression of epic narration. For traditions of shorter poetry, the difference in length had significant implications for variation in reproduction.22 Thus, the Old Norse short epic form was not ‘oral-formulaic’ in the manner of long epic forms. Like any other tradition, Oral-Formulaic Theory had to be adapted to approach the poetry on its own terms. The register of Old Norse poetry is characterized by formulaic language and formulaic strategies, even if their use when reproducing individual poems may 20 This difference in approach has consequences affecting what qualifies as a ‘formula’ as opposed to other constructions in an oral-poetic system. This includes conventional patterns of syntax (e.g. Mellor 1999), meter (e.g. Foley 1976) and collocative sets of words accomplishing alliteration or rhyme without producing a coherent unit of meaning (cf. Frog 2009: 236–239). 21 For an accessible overview, see Foley 1988; cf. Finnegan 1977: 52–87; in Old Norse poetries, see also Acker 1998: 85–110; Frog 2011a: 19–28 and works there cited. 22 See e.g. Holoka 1976: 572; Lönnroth 1971; Harvilahti 1992a; Lord 1995. It may be observed that the greatest obstacle in this discussion was that the Parry–Lord model had become an iconic image of Oral-Formulaic Theory, and this image was then popularly interpreted as a hegemonic model to be applied or rejected. 111 have been far less flexible than the epics studied by Parry and Albert Lord.23 In its development, Oral-Formulaic Theory remains concerned with the basic linguistic phenomenon of the formula24 as this is formally realized in relation to a vernacular poetic system and as it functions in relation to the tradition. When addressing traditions of shorter poetry, it is important to distinguish between the reproduction of socially recognized texts or textual entities25 and the production of new texts within (or in relation to) generic strategies and realized through the traditional register. It is equally important to recognize that the degree of conservatism or acceptable variation is not simply a function of meter, but also of convention in how poetry is presented and used (see e.g. Harvilahti 1992b). In addition, the crystallization of a textual entity will not necessarily be uniform – i.e. some sites in a text will exhibit less variation than others, even in South Slavic epic.26 Regarding the reproduction of textual enti- ties, formulaic language can potentially function mnemonically to reinforce conservative reproduction, and even where conventions of reproduction are quite variable, competence in the idiom plays an essential mnemonic role in the ability to learn new texts (cf. Rubin 1995; Harvilahti 2000). Irrespective of conventions of reproduction, the composition of new poetic text relies on competence in the idiom, including formulaic language. That competence is internalized through exposure to and participation in practice of the tra- dition. It is acknowledged that the skaldic “kenning is based on traditional paraphrasing patterns, and is not primarily the product of individual poetic inspiration” (Clunies Ross et al. 2012: lxxiii). Nevertheless, a persistent myth of scholarship is that Oral-Formulaic Theory has no relevance to skaldic poetry – with the implication that dróttkvætt poetry has no formulae comparable to those relevant to Oral-Formulaic Theory (cf. Birgisson 2007: 15). This myth is rooted in a misconception that Oral-Formulaic Theory and poetic formu- lae are only relevant to extremes of flexible composition in performance or 23 Cf. Lönnroth 1971; Harris 1983 [2002]; Acker 1998; Mellor 1999 [2008]; Haymes 2004; Thorvaldsen 2006; Frog 2011a; Þorgeirsson 2010; 2013. 24 However, the distinction of ‘formula’/‘not formula’ among the broad range of compositional resources does not always remain entirely clear (cf. Foley 1991). For those unfamiliar with Oral- Formulaic Theory, it should also be pointed out that Oral-Formulaic Theory is not exclusively concerned with verbal formulae, but also with constitutive content units represented through formulaic language as well as compositional wholes. 25 On textual entities, see Frog 2011b: 8–11. 26 E.g. Gil’ferding 1894: 24; Lord 1960: 58–63, 69–71; 1995: 22–62, 167–182; cf. also Foley 1991: 206; Siikala 1990: 84–85. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 112 Frog improvisation. Skaldic poetry was, in contrast, oriented to the production of unique texts and the conservative reproduction of these as socially circulated textual entities. However, the use of crystallized, metrically entangled diction for the production of well-formed poetic lines is not exclusive to composi- tion in performance.27 Internalizing the tradition in relation to exemplars was essential to both composition and the mnemonics of reproduction. The long- term continuities in conventions of meter and diction alone suggest that this poetry tradition maintained and communicated strategies and resources for meeting metrical demands when realizing these through the register.28 In order to introduce kennings as semantic formulae in skaldic verse, it is useful to begin from an extreme of crystallization that produces a more familiarly recognizable ‘formula’ and then to gradually advance into ranges of flexibility. First, however, it is necessary to point out that the statistical meth- ods of early Oral-Formulaic Theory produced the misleading impression that a formula is qualified by statistical demonstrability – i.e. because the research method demonstrated formulae statistically. It simultaneously suggested a formula was qualified by multiple contexts of occurrence – i.e. according to research questions about variation in reproduction, and research aims to dem- onstrate that particular texts were neither produced as works of written litera- ture nor transmitted as ‘memorized’ textual entities. Neither of these are valid criteria for determining formulaicity (Wray 2002: 25–26). These conditions had the significant consequence of excluding expressions conventionally used in only a single context of a single textual entity because these were aligned with memorization rather than composition in performance.29 Thus in studies of Old Norse poetry, independently attested lines and their variations in the same textual entity were not regarded as reflecting conventional formulae. 27 It may be observed that oral-formulaic systems can be maintained for the composition of new poetic texts even where the textual entities are conventionally circulated in written form (Blackburn 1988: esp. 24–25). 28 A broad range of evidence suggests that complex skaldic verses could be composed in situ, including different types of verse exchanges and challenges for skilled poets to compose on prescribed topics. This evidence suggests that the rapid, generative production of contextually appropriate and metrically well-formed text was part of social practice (Frog 2009; cf. Grove 2008). If this was the case, it implies conventional resources and strategies that would enable rapid composition as a socially and historically sustainable aspect of practice. 29 This same framework in fact could be used to disqualify repeated expressions only docu- mented in a single poem from being considered ‘formulae’ (Acker 1998: 91–94; cf. also Frog 2010b: 101–102). For indications of formulae in Germanic poetry traditions that only became acknowledged through cross-cultural comparison or historically remote evidence, see e.g. Kel- logg 1965; Þorgeirsson 2013: 351, 355–356, 363–366, 368–370. 113 However, Albert Lord (1995: 62) observes, “The larger the sample with which one works, the less adequate is the concept of word-for-word memorization as a means of song transmission.” Thus in the corpus of kalevalaic poetry, where there may be a hundred or more examples of a single textual entity, the size of the corpus motivates recognizing such context-specific lines and their variations in terms of formulae (cf. Frog 2010a: 400–405). Moreover, the conventional use of a formula in a single textual entity does not prevent it from being used elsewhere (Tarkka 2005: 65–67).30 When this is acknowl- edged, kennings in socially circulating verses of skaldic poetry can be regarded as verbally crystallized31 and metrically bound formulae. In addition, it was not always necessary or even motivated for everyone to reflect on why such a kenning meant e.g. ‘battle’ if the verses were transmitted with information relevant to their content or the referent of the kenning was otherwise formally and/or contextually transparent (see Wray 2002: 130–132; cf. Lindow 1985: 27–28). In other words, users familiar with verses and conventions of versi- fication need not necessarily analyze them in order to interpret a kenning appropriately (cf. Wray 2002: 131).32 Although it has not been customary to view such textualized kennings as formulae of the classic Parry–Lord type, this category of formula becomes relevant for consideration where verbally identical kennings are encountered realizing the same metric-structural basic 30 The number of compositions in which a verbal sequence is preserved should also not be conflated with the number of times particular verses are heard and used by individuals (i.e. its average token frequency). Where a formula is conventional to a single context, it might index that context referentially if it were recognizable (see Frog 2010b: 102–103). 31 In oral transmission, the degree of verbal fixity would always be open to degrees of vari- ation. In the basic type addressed below, for example, the base-words in even-line uses of the formula could, in theory, be freely interchangeable without affecting either meter or sense as long as they did not alliterate. 32 The contextual recognition of a familiar type of kenning construction would also enable someone fluent in the poetic idiom and its patterns of paradigmatic substitution to intuitively apprehend its significance without recourse to parsing it analytically according to patterns of kenning formation. A dróttkvætt poem of only thirteen stanzas would exceed one hundred lines. Continuous oral delivery of 13 stanzas can be estimated to minimally require 5–10 min- utes (N.B. – delivery could also be characterized by regular repetition at the level of some unit of composition such as a line, couplet or half-stanza). The faster the rhythm of the mode of expression, the greater the need for reliance on immediately apprehending conventional verbal constructions in order to follow the sequence of the composition. It may also be noted that a fairly high degree of apprehension by at least prestige members of an audience would seem to be historically required as a norm for the debut performance of longer oral poems to remain interesting, desirable and worthy of patronage in long-term social practice. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 114 Frog type, especially where these are found in full identical lines, as in examples (1a.i–v) and (2b) below.33 Linguistic formulae may potentially vary in the morphology of participat- ing lexical items. The base-word (NP1) of a kenning may vary morphologically while the determinant’s (NP2) morphology is dependent on the syntax of the construction (NP2-GEN NP1 or NP2-NP1). Nevertheless, the frequently-occur- ring Odin-heiti Yggr, for example, can receive alternative genitive markers /s/ and /jar/ (Ygg-s, Ygg-jar). These alternative markers are allomorphs – inter- changeable morphemes – that realize the morpheme-equivalent unit ‘Odin- GEN’. The choice of allomorph is metrically conditioned in dróttkvætt, varying in relation to the syllabic rhythm of the line in which it appears. Participating lexical items of a formula vary according to ‘slots’ in the construction. In other words, the formula is constituted of a construction in which, following Wray, certain lexical items have crystallized into a distinct, morpheme-equivalent expression. These slots will be syntactically and (at least to some degree) semantically conditioned in relation to the construction. Classes of slot-fillers might be as general as the subject or object of a verb as in X thinks nothing of Y, or they may be as narrowly idiomatic as to know/learn the ropes (Wray 2002: 50; Schmitt & Carter 2004: 5–7). In the latter case, these may come to qualify as distinct, crystallized formulae (also called ‘prefabs’) realizing a distinct unit of meaning, groups of which forming what Parry (1930: 85) called ‘systems’ of formulae (cf. Foley 1988: 28–33; Acker 1998: 39–43). A still narrower category of open-slot formula realizes a consistent unit of meaning, such as a piece/slice of the action (Wray 2002: 50). In most discourses, variation in such an idiom is most likely attributable to dialect, because alternate realizations of such con- ventionalized expressions appear in general to “repel one another”: a compe- tent user will not normally alternate between them (Kuusi 1975: 59; although cf. Frog 2010b: 100). In oral poetry, open slots will also be metrically condi- tioned. Thus an oral-poetic register that has developed in relation to phonic and/or syllabic metrical requirements will normally maintain archaisms and be enriched by dialectal words as well as foreign loans that facilitate realizing the meter and associated stylistic devices such as parallelism (Coleman 1999: 38; Foley 1999: 74–83). In addition, “[p]honic demands have semantic impli- cations” (Roper 2012: 83), stretching the semantics of terms and subordinat- ing them to synonymy (ibid.), or employing formulaic elements for purely 33 It should also be born in mind that the preserved corpus of skaldic poetry reflects only a fraction of the verse that was composed and circulated in the various regions of Scandinavia across a period of several centuries. 115 metrical reasons with little or no semantic weight.34 Albert Lord (1960: 48) drew attention to a formula zasedit X = ‘to mount X’, in which the verb varied morphologically in relation to the subject and the formula could be realized with 13 different two-syllable words for ‘horse’. The completed formula invari- ably realized a coherent unit of meaning – i.e. the formula ‘to mount X’ was in fact ‘to mount a horse’. Lord observes that the diversity of terms realizing an otherwise stable formula of this type is attributable to pragmatic variation according to the “acoustical context” while maintaining “the same essential meaning and metrical value” (ibid.: 53). When the formula is approached as a morpheme-equivalent unit, variation between semantically equivalent lexemes according to the metrical environment is equivalent to the variation between Ygg-s and Ygg-jar mentioned above. The 13 two-syllable terms for ‘horse’ can therefore be considered allomorphs in realizing the formula. Old Norse heiti are conventionally approached according to categories of semantic equivalence.35 Heiti present equivalent sets of elements in the poetic register that have undergone semantic levelling to a degree that paradigmatic substitution in kenning formation can vary pragmatically according to the metrical environment. They can therefore be approached as registral allo- morphs, or variants of the same morpheme. This does not mean that variation between allomorphs cannot be semantically significant or indexically marked in individual cases (even in use of Ygg-jar as opposed to Ygg-s). However, it allows heiti to vary pragmatically, especially where these are used to realize a formulaic unit of meaning to which the heiti may be wholly subordinated (cf. Frog 2011a: 32). Such pragmatic variation appears to have been fundamental to compositional practice. Examples (1–2) below present metrically entangled kenning formulae directly comparable to Lord’s example from South Slavic epic above: one element of the kenning is stable and the second element is in variation realizing a single semantic unit with a consistent metrical value (dynr/gnýr [NP2-GEN] = battle in the final three positions of a line). These illustrate the same phenomenon in dróttkvætt poetry. Defining the formula 34 The classic example is Parry’s (1928) discussion of ‘swift-footed Achilles’, in which the epithet ‘swift-footed’ can appear for purely formal reasons of the meter and the whole formula simply means ‘Achilles’; for corresponding examples from Old Norse eddic poetry, see e.g. Frog, Roper 2011; Frog 2011a: 33; for examples of alliteration as a determinant on lexical choice or formula in eddic poetry, see e.g. Mellor 1999: 119–122; Lönnroth 2002: 17; Acker 1998: 64–65; Frog 2011a. 35 This has not always been the case: scholarship has been characterized by a long-standing inclination to approach individual heiti as “never completely interchangeable” (Frank 1978: 41), particularly where aesthetics of verses as unique works are emphasized and traditionality of diction is marginalized. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 116 Frog according to morpheme-equivalence emphases not only continuity in a unit of meaning but also in signifying elements as well. In kennings, however, a metrically entangled kenning can vary both lexical constituents at the level of surface texture without altering the semantic unit being realized. Rather than one element of the formula being stable and the other an allomorph, elements in both metrical positions may be realized according to allomor- phic variation. In this case, morpheme-equivalence can be considered to be maintained through allomorphic variation according to contextual phonic demands conditioned by constraints of the number and weight of syllables for the metrical positions with which the formula is entangled. This is described as a semantic formula – a formula that corresponds to a morpheme-equivalent unit although all of its elements may exhibit allomorphic variation (cf. Holland 2005). The approach to the oral-poetic formula through metrical entangle- ment rather than Parry’s (and Lord’s) metrical boundedness results in viewing such kennings as a linguistic phenomenon of (registral) semantic formulae that have become metrically entangled through the social historical condi- tioning of language use in the tradition. Consequently, metrically entangled kennings characterized by the same semantic formula should not be regarded as independent of one another but rather as metrically entangled reflexes of a common semantic formula of the register. This leads to the question of the degree to which the allomorphic constituents of battle-kenning formulae were restricted or open in generative production and whether battle-kennings are characterized by one or several semantic formulae. Contrary to expectation, the pilot study revealed that the allomorphic elements realizing metric-structural basic types of battle-kenning were not bound by semantic equivalence class. Moreover, the relative frequency of both base-words and determinants could be correlated especially with patterns of alliteration and secondarily by capacity for rhyme. This suggested that battle- kennings realize a relatively coherent and flexible semantic formula in one of two syntactic constructions (NP2-GEN NP1 or NP2-NP1 = ‘battle’) rather than a set of distinct semantic formulae all meaning ‘battle’.36 Some exceptional 36 In other words, ‘(weather) (of weapons)’, ‘(weather) (of a valkyrie)’, ‘(meeting) (of imple- ments of battle)’, ‘(sound) (of a mythic agent of battle)’, and so forth were not exhibited as distinct semantic formulae within the data at the level individual metric-structural types. This does not reduce the value and relevance of different typologies of kennings based on distin- guishing such categories, but it suggests that the differentiation of semantic equivalence class could be subordinated to the pragmatics of the metrical context in which the kenning was realized. This finding cannot be generalized to all kennings. For example, kennings for the god ‘Thor’ reflecting kinship relations, such as ‘son of his mother’, are clearly distinct from kennings reflecting his relationship to adversaries, such as ‘slayer of a monster’ (otherwise generative 117 words appeared in variations on the formula that conditioned the seman- tic class of the co-occurring heiti (cf. 5.i–iii below). This approach does not exclude more general conventions of preferred association between certain semantic categories (cf. Meissner 1921), which requires further study. The present series of case studies explores the functioning of equivalence classes in relation to kennings as semantic formulae and the metrical entanglement of verbal elements where battle-kennings are conventional to certain metrical positions. These case studies will offer perspectives on the degree to which metric-structural types must be taken into consideration before attempting to assess more general conventions of association between equivalence classes of heiti because of their potential to skew data (cf. 1a.i–v, 1a.xvii and 2b below). The dróttkvætt Meter Among the oral poetries of Europe, dróttkvætt is an exceptionally complex and demanding meter. In practice, the meter had a number of conventional variations with slight differences in metrical constraints, although the basic features remained generally consistent. Dróttkvætt was essentially a syllabic meter (with rule-governed flexibility) composed in couplets, with two cou- plets forming a half-stanza called a helming and a total of eight lines forming a stanza. Its rules of syntax (see Gade 1995) allowed a remarkable scrambling of language across the four lines (two couplets) of a helming, and two (rarely more) clauses or independent statements could be interwoven across those four lines (cf. Wills 2009). The main conventional constraints relevant to the present discussion are rhyme, alliteration and syllable weight. Although it is not relevant here to open the details (and the debate) of qualifying ‘heavy’ and ‘light’ syllables, a syllabic quality referred to metaphorically as ‘weight’ is a factor which prefers or prevents certain syllables from occurring in certain metrical positions (see Kuhn 1983; Árnason 1991; Gade 1995). Although the meter is conventionally recognized as three metrically stressed positions, the relationship between metrical stress, syllabic weight and word stress or phrasal stress is unresolved and will not be debated here. For the present discussion, it suffices to observe that conventions of syllabic weight condition the fourth and variation would produce expressions like ‘son of a monster’ and ‘slayer of his mother’ = Thor). Nevertheless, the data of the pilot study suggests that the paradigms according to which ken- nings exhibit variation are not necessarily limited to individual equivalence classes of heiti, hence paradigmatic substitution may occur within sets of equivalence classes conventional to the kenning as a semantic formula. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 118 Frog fifth positions of the six-position line (the final position being an unstressed inflectional ending). This results in constraints on which heiti can be used in realizing kennings of certain metric-structural types. Two lexically stressed syllables in odd lines should alliterate.37 In even lines, the first position should share this pattern of alliteration, uniting the couplet, but not any of the following lexically stressed positions. The penultimate strong position in a line should carry a syllable that rhymes with a preceding lexically stressed syllable in the same line. In odd lines, the rhyme should (normally) not include the syllable’s vowel, a type of rhyme called skothending (e.g. 3b.ii below: lind sprakk í rym randa). In even lines, the vowel should participate in the rhyme, a type of rhyme called aðalhending (e.g. 2a.i below: hjalmstall í gný malma). However, dróttkvætt variations had different degrees of flex- ibility. The precise requirements of rhyme varied and rhyme was sometimes neglected (e.g. 1b): rhyme was generally secondary to alliteration in conven- tional requirements of dróttkvætt composition. Most important to recognize is that odd and even lines thus present slightly different metrical conditions. The difference in these conditions must be taken into consideration when observing formulaic language which can develop distinct conventions for each of these two metrical environments. 'Mapping' Metrically Entangled Constructions in dróttkvætt The term basic type is used to designate the distribution of ‘simple’ kennings, consisting of only a base-word and a single determinant, across the metrical positions in a line, couplet or helming. Complex types consisted of three or more elements (complex kennings). Within the pilot study, complementary adjectives, numbers and articles were, in general, found to be freely distribut- able through a stanza surrounding a ‘simple’ kenning, without structural or semantic effect on the kenning as such. For the present study, such examples are considered to complement basic types but not to produce variations on 37 It is common to presume that alliteration in dróttkvætt should be identified with metrically stressed positions on analogy to other Germanic alliterative meters which have a stressed-based rhythm (ultimately following Sievers 1893). However, dróttkvætt is characterized first and fore- most by a rule-governed syllabic rhythm, and it remains unclear how conventions of stress-based rhythms may have been integrated into it or subordinated to syllabic-based rhythms. It thus warrants observation that the adjacent Finnic cultures also had an alliterative syllable-counting meter with quantity-based rules for stressed syllables, and those rules in fact conventionally require certain alliterating syllables to occur in metrically unstressed positions (cf. Frog 2010). 119 the basic type. Adjectives are only seen as producing variations on basic types in cases where this affects the position of the base-word or the determinant in the manner of developing a complex kenning. Complex kennings appeared as extensions or expansions of basic types by combining two basic types that have intersecting metrical positions. The details of this are not relevant here because such variations are not included in the examples surveyed here, although a few will be mentioned in the associated discussion. For the purposes of this study, constructions are ‘mapped’ in the six metrical positions of lines with the simple system used in the pilot study. The six-position sequence is represented numerically for each line (123456). Positions filled with a kenning’s base-word are replaced with an ‘X’ and those of the determinant with a ‘Y’. Uppercase characters represent the stressed onset syllable and lowercase characters represent unstressed positions. In a few examples for discussion, complex kennings can be extended with ‘Z’ representing the determinant in the determinant kenning. Thus the battle-kenning sverða glaumr = ‘racket of swords’ in the line glaum-herðǫndum sverða (Kolli Ingdr 3II.6) = ‘racket-strengtheners of swords’ = WARRIORS is mapped X234Yy. The warrior-kenning in the same line is mapped YXxxZz. When relevant, ‘p’ indicates a preposition, and is placed in parentheses where this is presented as conventional but subject to variation (cf. 2a below). In cases where metrically entangled words or strings are observed, these will be separated by hyphens. Thus, depending on the emphasis, the for- mula in the line Eiríkr í dyn geira = ‘Eiríkr in the din of spears’ (HSt Rst 22I.6; ÞHjalt Lv 1I.8) can be mapped 12-í-dyn-geira, 12-í-dyn-Yy, 123-dyn-56, 12pXYy, 12(p)XYy or 123XYy. The basic types reviewed here are all situated within a met- rical line. Variations which involve an element in an adjacent line will indicate the line-break with ‘/’. Although the detailed mapping of types is less essential for the opening case study, it will become increasingly important as different types are placed in relation to one another. Dynr in Basic Type 12(p)XYy As observed above, it is practical to begin approaching metrically entangled semantic formulae through examples exhibiting pronounced verbal crystal- lization, producing formulae in a sense that is widely familiar from Parry’s definition. The verbal fixity enables the formula to be observable at a rather superficial surface level of the text, which makes this type of example an accessible point of departure for considering more complex phenomena. In example (1), the noun dynr = ‘din’ provides a base-word in a battle-kenning construction. In dróttkvætt poetry, the simplex noun dynr (i.e. not used in a Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 120 Frog compound word) appears invariably as the base-word in battle-kennings and is found almost exclusively in a single metrical position, and more specifically in battle-kennings of the metric-structural type 12(p)XYy and its variations, all but one of which are found in even lines (Frog 2009: 234–236). The basic formula is the phrase í dyn = ‘in the din’ followed by a two-syllable determinant that forms a battle-kenning (í dyn [NP2-GEN] = ‘in battle’). Together, these fill the last four syllables of a line. The formula has a simpler form or variation without the preposition. The near-exclusive attested use of the simplex dynr in a single metrical position and in regular constructions makes formulaic use statistically unambiguous. Examples in even and odd lines are separated here for easier ref- erence in later discussion. The two variations on the basic type (Frog 2009: 236) are not relevant to the present discussion and are not included here. The near-exclusive use of dynr in battle-kennings of this metric-structural type is an unambiguous relevant indicator that this is a ‘formula’ in a classic Parry–Lord sense. The use of dynr in this construction is itself metrically conditioned. The determinant in position 4 must be ‘light’. This means that if it has a short vowel (dynr), the vowel cannot be followed by a consonant cluster (inflected dyn, but not inflected dyns = ‘of the din’). If it has a long vowel (gnýr = ‘roar’), this cannot be followed by a consonant (inflected gný but not inflected gnýs = ‘of the roar’). Consequently, this battle-kenning basic type cannot form a determinant of a com- plex kenning. Gnýr and dynr are both heiti for ‘noise’. Although the correspond- ing conditions in other basic types may prevent the use of dynr in particular contexts, this does not explain its limited range of use in the poetry, especially in contrast to the far more common gnýr. Significantly, the restricted use of dynr is an indicator that this lexical item has become metrically entangled with a specific metric-structural type of battle-kenning: it was not employed as a base-word in battle-kennings with a more even distribution across basic types. In other words, the use of dynr was not freely generative in versification although it was otherwise a common noun rather than a strictly poetic term (cf. Frog 2012: 58–60). The formulaic use of dynr is a useful case to begin with when considering potential cases of metrical entanglement involving terms or classes of equivalent terms that may otherwise also be employed with a greater range of uses in the corpus. In this regard, it may be observed that 5 of the 19 examples, or more than 25%, are realized with the determinant geirr = ‘spear’ (1a.i–v). This could be a relevant indicator of a degree of metrical entanglement leading to conventional preferences in word-choice when realizing the formula. This would mean that an associative index developed and became socially established between this battle-kenning basic type and certain lexical items in compositional practice. However, in this case, 3 of the examples may reflect conventionalized expres- sions at the level of the full line Eiríkr í dyn geira (1a.i–iii, used with reference to 121 different kings of the same name). This suggests the possibility of crystallization at the level of a full line. The prominence of geirr in this data may thus be at least partially dependent on its association with a crystallized line. Table 1a. 12-í-dyn-Yy in even lines38 39 40 12(p) X Yy Translation of (p)XYy38 Source Eiríkr í dyn geira ‘...in the din of spears1’ HSt Rst 22 I.6 Eiríkr í dyn geira ‘...in the din of spears1’ ÞHjalt Lv 1 I.8 Eirekr í dyn geira ‘...in the din of spears1’ Bbreiðv Lv 6 V.4 ófár í dyn geira ‘...in the din of spears1’ Anon Krm 18 VIII.6 meir framm í dyn geira ‘...in the din of spears1’ GunnHám Lv 10 V.6 Sigurðr fell í dyn vigra ‘...in the din of spears2’ Anon (Nj) 3 V.6 morðteins í dyn fleina ‘...in the din of shafts’ KormǪ Sigdr 2III.2 einarðr í dyn sverða ‘...in the din of swords1’ Anon Krm 23 VIII.8 hné ferð í dyn sverða ‘...in the din of swords1’ Ingj Brandfl 3 IV.2 skǫrungr fell í dyn hjǫrva ‘...in the din of swords2’ Ingj Brandfl 4 IV.2 hjalmskúr í dyn malma ‘...in the din of metals’ RvHbreiðm Hl 58III.2 almskúr ok dyn hjalma ‘...and the din of helmets1’ Þtréf Lv 1 IV.4 malmþings í dyn hjalma ‘...in the din of helmets1’ Skúli Svǫlðr 2 III.6 malmr gnast í dyn hjalma ‘...in the din of helmets1’ Anon (Nj) 3 V.4 folkvaldr í dyn skjalda ‘...in the din of shields1’ ESk Geisl 14 VII.6 hvít, svǫrt, í dyn rítar ‘...in the din of the shield2’ RvHbreiðm Hl 39 III.8 lét ek sunnr í dyn Gunnar39 ‘...in the din of Gunnr (valkyrie)’ Hfr Lv 13V.6 bryngǫgl í dyn Skǫglar40 ‘...in the din of Skǫgul valkyrie)’ Þhorn Gldr 5I.4 Table 1b. 12-í-dyn-Yy in odd lines 12(p) X Yy Translation of (p)XYy Source áðr létti dyn darra ‘... the din of darts’ Bjbp Jóms 39I.7 38 Translations are presented only of the semantic formula. The flexibility of syntax in the poetry tradition makes the formula’s relationship to other elements in the same line highly variable. These words may participate in a different clause or sentence (indicated by ‘–’ in examples 2a.vii and 2a.x below). Translating these and explaining the kennings which in some cases occur would only distract from the purpose of the examples. 39 One manuscript reads for this line: lét ek baldr riðins skjaldar. 40 This line reads “bryngaul ara Skǫglar” in the Flateyjarbók manuscript. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 122 Frog It may initially be observed that the different metrical conditions of odd and even lines have implications for the realization of the basic type. The battle-kenning is realized in positions 4–6 in the line. In even lines, alliteration should occur only in position 1, dictating that neither the base-word nor the determinant carry alliteration (1a), whereas both elements carry alliteration in the one odd-line example (1b). The situation of the determinant in positions 5–6 requires that it carry the rhyme in the line. The unambiguous predomi- nance of even-line use highlights a potential value for meeting aðalhending- rhyme – i.e. rhyme in which the vowel participates. Variation in determinants is not constrained to a narrow semantic field. Even if heiti for ‘weapons’ (1a.i– xi, 1b) and for ‘armour’ (1a.xii–xvi) are considered a broad category of ‘imple- ments of battle’, the valkyrie-names (1a.xvii–xviii) unquestionably belong to a distinct category of heiti. Although only 2 of the 19 examples employ names associated with mythology, these appear to be involved in an otherwise regu- lar generative formulaic model. Within that model, these names are engaged with the metrical requirement of rhyme no less than words like geirr. Before opening a discussion of the role of these words in rhyme, it is practical to first introduce other examples of this basic metric-structural type of battle kenning with different base-words and discuss their use as determinants in that context. Gnýr in Basic Type 12(p)XYy Basic type 12(p)XYy also exhibits a formula of the Parry–Lord variety corre- sponding to that with dynr but realized with gnýr = ‘roar’, which is much more widely and flexibly employed in the poetic register of dróttkvætt. This formula is also found with greater frequency and greater variation, including literal uses of 12-í-gný-[NP-GEN] without forming a kenning (e.g. Hallm Hallkv 1V.7–8; Anon (Mberf ) 7II.7). The data presented in example (2) is considered sufficient to illustrate that use is formulaic. The flexibility of this formula and the range of variations on basic type 12(p)XYy are not relevant to the present discussion and will not be explored here. Examples of formulaic uses of dynr and gnýr illustrate metrically entangled or potentially bound formulae with conventionally established base-words and variable determinants forming battle-kennings. Use of both formulae is pre- dominantly in even lines, even if gnýr exhibits greater flexibility in this regard. Even-line use situates kennings of this type outside of alliteration requirements and only the determinant exhibits participation in rhyme. The diversity of deter- minants across these examples illustrates the variability of heiti in composition. It also highlights that variation must be considered in dialogue with the role of 123 rhyme as a determinant on word-choice in generating a metrically well-formed line. In other words, the base-word of basic type 12(p)XYy is not subject to needs of meeting phonic requirements (as opposed to e.g. type 1(p)YyXx). In odd lines, the requirement that alliteration be carried by two lexically stressed syllables makes it probable that at least one element of the kenning, comprised of nouns, will alliterate. In odd lines, the base-word still does not participate in rhyme, but both the base-word and determinant carry alliteration in the line. Table 2a. 12(p)-gný-Yy in even lines 12(p) X Yy Translation of (p)XYy Source lands folk í gný branda ‘...in the roar of brands (swords)’ Ótt Hfl 9I.6 fjǫrspell í gný hjǫrva ‘...in the roar of swords2’ Þfagr Sveinn 1 II.2 sigrtrúr í gný vigra ‘...in the din of spears2’ Anon Líkn 39 VII.4 saddr varð í gný nadda ‘...in the roar of barbs’ HaukrV Ísldr 5IV.6 herr œxti gný darra ‘... the roar of darts’ Bjbp Jóms 30I.2 Áláf, í gný stála ‘...in the roar of steel’ ESk Geisl 54VII.4 málgrǫ́ðr – í gný stála ‘...in the roar of steel’ GunnlI Lv 10V.8 þokum framm í gný stála ‘...in the roar of steel’ Rv Lv 31II.6 hjalmstall í gný malma ‘...in the roar of metal’ Hharð Lv 14II.8 almþing – í gný malma ‘...in the roar of metal’ Þorm Þorgdr 10IV.6 Þórbjǫrn í gný fjǫrnis ‘...in the roar of a helmet2’ HaukrV Ísldr 17 IV.8 aldrœn í gný skjalda ‘...in the roar of shields1’ SturlaS Lv 1 IV.2 hug-Baldr í gný skjalda ‘...in the roar of shields1’ Hfr Lv 13 V.2 benvǫnd í gný randa ‘...in the roar of borders (shields)’ ÞBrún Lv 4V.2 Table 2b. 12(p)-gný-Yy in odd lines 12 (p) X Yy Translation of (p)XYy Source þar varð í gný geira ‘...in the roar of spears1’ Glúmr Gráf 5 I.3 jafn vas mér í gný geira ‘...in the roar of spears1’ Egill Lv 34 V.7 ok tveir at gný geira ‘...at the roar of spears1’ Tindr Lv 2 V.3 en sunnr at gný Gunnar ‘...at the roar of Gunnr (valkyrie)’ Hókr Eirfl 8I.5 hinn’s fór í gný Gunnar ‘...in the roar of Gunnr (valkyrie)’ Eskál Vell 33I.3 frák hann við gný Gunnar ‘...at the roar of Gunnr (valkyrie)’ RvHbreiðm Hl 65III.3 vér unnum gný Gunnar ‘... the roar of Gunnr (valkyrie)’ Ólhelg Lv 6I.3 Einn háði gný Gunnar ‘... the roar of Gunnr (valkyrie)’ Refr Giz 1III.1 Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 124 Frog The difference in odd-line and even-line distribution of the two base-words can be considered as potentially related to the relative frequency of different patterns of alliteration, among which d-alliteration is rare while g-alliteration is fairly frequent in Old Norse poetry (cf. Hollmerus 1936). It is therefore striking that dynr, otherwise rare in dróttkvætt, is better attested in even-lines (without constraints of alliteration or rhyme) than gnýr, which is otherwise common in the register. However, alliteration may in some cases have functioned as a nega- tive factor in the sense that dynr or gnýr could not occur with the determinant carrying rhyme in an even line without creating an alliteration that should be avoided. This is only a potential factor in one case with gnýr, where dynr would otherwise alliterate with the determinant darr = ‘dart’ (2a.xiv). However, this is found in 6 cases with dynr, where otherwise gnýr would alliterate with geirr = ‘spear’ (1a.i–v) or with the valkyrie-name Gunnr (1a.xvii), as well as an additional potential case where gnýr would produce an undesired allitera- tion with the verb in the line (1a.xiv). This suggests that the choice of formula in even lines when realizing the metric-structural basic type may have been conditioned by the avoidance of alliteration with the kenning determinant.41 Odd-line uses of gnýr are suggestive of the possibility that the kennings geira gnýr (2b.i–iii) and Gunnar gnýr (2b.iv–viii) may have been fully con- ventionalized expressions realizing the kenning in this basic type. Thus the semantic formula crystallized at the lexical level into an open-slot formula, and the open-slot formula potentially crystallized into distinct, conventional for- mulae, while the relationship of í dyn geira to Eiríkr í dyn geira above (1a.i–v) can be viewed as a similar process at the level of a full line. In addition to the 5 examples of the kenning Gunnar gnýr in the basic type, there is also an extended variation: Gunnar gagls gnýr in the couplet hnigu menn í gný Gunnar / gagls fyr strengjar hagli (ESk Ingdr 3II.5–6) = ‘the clash of the gosling of Gunnr before the hail of the bowstring’. ‘Gunnr’s gosling’ is a kenning for ‘bird of carrion’ (eagle or raven) and ‘roar of the bird of carrion’ forms a battle- kenning.42 The complex kenning does not affect the form of the basic (í) gný 41 Corresponding alternation of semantically and functionally equivalent open-slot formulae according to alliteration is attested for eddic narrative poetry (Acker 1998: 64–65; Frog 2011a: 40–44). In the use of Gunnr with dynr, it may also be observed that rhyme with sunnr is also found in (2b.iv), which exhibits aðalhending rhyme although it is an odd line, and Gunnr is also used in aðalhending rhyme in (2b.vii). 42 Cf. Meissner (1921: 147, 190), who seems to interpret this not as ‘raven’ but as ‘arrow’, which is perhaps an attempt to confer more logic on the kenning than it warrants and is certainly influenced by the alternative reading of the even line as strengjar gagls hagl = ‘hail of the gos- ling of the string’. Both battle-kennings would be complete independent of gagl, which is not a semantic improvement to either of them. 125 Gunnar kenning: gagl is outside of that structure and the 5 examples make it clear that the kenning would be semantically complete without the additional element (cf. Noreen 1921: 27–28; Sverdlov 2003). The term gagl carries allitera- tion at the onset of the following line and also carries the rhyme of that line. The number of additional examples of the Gunnar gnýr kenning realizing this metric-structural type is itself exceptional in the corpus (cf. Björnsson –2001). It is therefore highly improbable that a sixth example as a complex kenning is independently compositional. This kenning can most reasonably be regarded as a variation on a conventional, verbally crystallized formula (Gunnar gnýr → Gunnar gagls gnýr)43 that appears to be motivated by phonic demands of the meter as a strategy for meeting them. It can therefore be seen as a sixth example of Gunnar gnýr of the 12(p)XYy metric-structural type. The example is also of interest because the determinant ‘bird of carrion’ is otherwise only attested in battle kennings with the determinant ‘joy’ (cf. Meissner 1921: 201; Björnsson –2001). The equivalence class of determinants in other examples of this basic type varies between a broad category of ‘implements of battle’ and the category ‘valkyrie’ or more broadly ‘mythic agent of battle’. This example varies a verbally crystallized formula with the determinant ‘valkyrie’ by pro- ducing a complex kenning in the process of meeting metrical requirements. The formation of the complex kenning alters the class of determinant to a third, distinct semantic category of ‘birds of carrion’ or more broadly ‘beasts of the battlefield’ (Meissner 1921: 116–119).44 The fact that this is accomplished without formally altering the verbally crystallized formula highlights the fluid- ity of equivalence classes of heiti in compositional practice. When evidence of the 12(p)-gný-Yy formula is brought into comparison with examples of the 12-í-dyn-Yy formula, the prominence of geirr as a determinant (1a.i–v, 2b.i–iii) should be considered as at least potentially related to its allitera- tive use with gnýr. This presents potential for interference in preferential word- choice according to the basic type, a variety of metrical entanglement that will be explored in the subsequent case studies. The use of the valkyrie-name Gunnr is found only once in the 12-í-dyn-Yy formula (1a.xvii) and is therefore more 43 12(p)XYy + 1234Yy/X23456 → 12(p)XZz/Y23456. 44 It may be noted that the variation allows Gunnr to be interpreted as a common noun rather than as a proper name (i.e. gunnar gagl = ‘gosling of battle’ = bird of carrion) without violat- ing the rhetorical figure of the kenning (see Meissner 1921: 73–74, 201–202). Insofar as the use of Gunnr/gunnr in this kenning is inextricable from the kenning Gunnar gnýr, this example highlights the possibility that semantic ambiguity need not have been resolved in the production and interpretation of kennings. Additional examples of this strategy of varying basic types in relation to metrical demands will be observed in subsequent case studies in this series. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 126 Frog ambiguous in this regard. However, questions concerning the prominence of this valkyrie-name will be returned to at the conclusion of the present paper. Other Determinants in Basic Type 12(p)XYy Whereas 12-í-dyn-Yy and 12(p)-gný-Yy can be considered conventional for- mulae in their own rights, several additional examples of basic type 12(p)XYy are attested with sharply contrasting low frequency. As observed above, posi- tion 4 is metrically conditioned, requiring that the base-word be a ‘light’ mon- osyllable. This has a consequence that the majority of high-frequency base- words otherwise used in battle-kennings (e.g. hríð, él) cannot be employed in this position. Example set (3) presents variants with a base-word in the semantic equivalence class of ‘noise’ to which dynr and gnýr belong. Table 3a. 12(p)XYy with ‘noise’-heiti base-words in even lines 12 (p) X Yy Translation of (p)XYy Source siðfornir, glym járna ‘... the crash of iron’ Bjbp Jóms 7I.6 Table 3b. 12(p)XYy with ‘noise’-heiti base-words in odd lines 12 (p) X Yy Translation of (p)XYy Source sveit varð í rym rítar ‘...in the noise of the shield2’ ESk Elfv 2II.7 lind sprakk í rym randa ‘...in the noise of borders (shields)’ Eviðs Lv 5V (Heið 16).3 þrǫng at rym randa ‘...at the noise of borders (shields)’ ÞGísl Búdr 3I.3 verit hefr í þrym þremja ‘...in the thunder of swords3’ KormǪ Lv 30V.7 bitu sverð í hlym rítar ‘...in the clash of the shield2’ Skáldh Branddr 5IV.4 Hnigu fjandr at glym Gǫndlar ‘...at the crash of Gǫndul (valkyrie)’ Anon (SnE) 6III.1 Immediately observable is a shift in predominance from even-line examples employing dynr and gnýr to odd-line examples. Odd-line realizations of basic type 12(p)XYy battle-kennings exhibit integrated conventions of alliterating the base-word and determinant. Verbal variation in these examples can be directly associated in almost all cases with not simply meeting the alliteration 127 requirement of the meter, but with meeting that requirement according to the conventions of the metrically entangled formula. In addition, it is rather striking that all of the base-words in this group have a stem containing -rym- or -lym-, and all appear to be particular to the poetic register (although þrymr also appears as an epithet attached to personal names and as the name of a giant in the eddic poem Þrymskviða). If additional semantic fields of heiti are included for base-words, more exam- ples of 12(p)XYy battle-kennings can be observed. For example, the term byrr = ‘breeze’ is a weather-heiti which can function as a base-word for battle-kennings (4a–b). The simplex inflectional form byr seems only to occur in the position appropriate to this metrical-structural type of battle-kenning. Although infre- quent, this base-word exhibits even distribution across odd and even lines. Table 4a. 12-í-byr-Yy in even lines 12 (p) X Yy Translation of (p)XYy Source ítr varð í byr rítar ‘...in the breeze of the shield2’ Þórðh Lv 6 V.6 vápnsóttr í byr Þróttar ‘...in the breeze of Þróttr (Odin)’ Grett Lv 21V.2 Table 4b. 12-í-byr-Yy in odd lines 12 (p) X Yy Translation of (p)XYy Source nema rǫnd í byr branda ‘...in the breeze of brands (swords)’ ESk Geisl 53VII.5 hinn’r skyndi byr branda ‘...in the breeze of brands (swords)’ RvHbreiðm Hl 59III.7 Other base-words used in corresponding expressions include semantic fields that condition or limit the determinants: svipr = ‘a swoop’ can only form a battle-kenning with terms for hand-held weapons whereas flug = ‘a flight’ forms a battle-kenning with terms for weapons that can be used as projectiles (5). These base-words can be considered variations on the battle-kenning as a broad semantic formula that do not allow free combination with determinant heiti. However, Meissner (1921: 192) identifies one example of the valkyrie- name Gunnr as a determinant for svipr, suggesting flexibility even here. Table 5. 12(p)XYy with other base-words in odd lines 12 (p) X Yy Translation of (p)XYy Source þars gerðum svip sverða ‘... the swoop of swords1’ Hharð Gamv 5 II.3 harðr rauð í svip sverða ‘...in the swoop of swords1’ RvHbreiðm Hl 65 III.5 en mín at flug fleina ‘...at the flight of shafts’ Sigv Nesv 5I.5 Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 128 Frog Taken in isolation, the less frequent base-words do not stand out in the data. Taken together, a pattern is exhibited suggestive of a metric-structural construc- tion for realizing kennings through an extensive lexicon of possible heiti. Among these kennings, gnýr (22 examples) and dynr (19 examples) stand at the centre of potential base-words. The significantly lower frequency of other base-words (14 total examples) makes it possible that additional metrically and semantically viable heiti may also have been occasionally used but have not been observed owing to limitations of the data. If this were the only identified metrically bound kenning formula of this sort, the relationship between the alternative base-words would remain highly ambiguous, as would the degree to which this type of met- rically bound semantic formula should be seen as the exception rather than a fundamental of skaldic compositional practice. 12(p)XYy does appear to be exceptional – exceptional in the marked predominance of two base-words with which it is primarily realized: the base-words dynr and gnýr account for 41 of the 55 examples of the basic type that have been identified (without including variations). This is approximately 75%, which, on the basis of the pilot study, appears to be quite exceptional among variation observable in different basic types. On the other hand, this exceptional quality makes it easier to approach the principles of metrically entangled kenning formulae in later parts of this discussion where basic types exhibit variation in both the determinant and also in the base-word. The patterns beneath the surface of lexical variation might not otherwise be immediately transparent to readers less familiar with skaldic dróttkvætt poetry. Noise of Swords, Noise of a Valkyrie In the 55 examples listed above, names from the mythology appear as deter- minants in 9 cases, which is approximately one in six or a bit more than 16%. These are the three valkyrie names Gunnr (1a.xvii, 2b.iv–viii), Gǫndul (3b.vi) and Skǫgul (1a.xviii), and one example of the Odin-heiti Þróttar (4a.ii). The 5 examples of Gunnar gnýr and the sixth variation in a complex ken- ning were hypothesized above to reflect a conventionally crystallized expres- sion. This finds support in the fact that the other three odd-line uses of gnýr in the data all have the determinant geirr (2b.i–iii), two of the three odd-line uses of rymr have the alliterative determinant randr (3b.ii–iii), both odd-line uses of byrr carry alliteration with brandr (4b.i–ii), and both uses of svipr do the same with sverðr (5.i–ii). Of the 19 examples of odd-line uses of the 12(p)XYy, only 5 do not exhibit a recurring pairing of base-word and determinant (1b, 3b.iii– v, 5.iii). Of the 5 exceptions, 4 are only attested once and may therefore only 129 appear to be exceptional owing to limitations of the data. The general pattern in the data suggests crystallization in the pairing of alliterative determinants with base-words in odd-line use for the accomplishment of alliteration. In the examples of 12(p)XYy, the determinant of the kenning always forms a final troche and the stressed syllable of this word always carries the rhyme of the line. This makes it reasonable to consider that lexical choice is condi- tioned by phonic demands in composition. In odd lines, the valkyrie-name Gunnr rhymes in -nn-. Following the index of Eysteinn Björnsson (–2001), the attested heiti used in simple battle-kennings exhibit three other words accomplishing rhyme in -nn-. Two of these three function as base-words in battle-kennings (senna = ‘insult exchange’, renna = ‘rush’) and are therefore not relevant to this metric-structural type. The only other determinant listed with this rhyme is unnr = ‘sword’ (or the Odin-heiti Unnr), where the ken- ning is the result of editorial emendation of the manuscript text (Þorb Lv 2V; cf. Jónsson 1967 A2: 449; 1967 B2: 481). Rhyme in -nn- on a determinant in battle-kennings was therefore conventionally accomplished with Gunnr or extension into a complex kenning would be required. This increases the probability that use of Gunnr as a determinant and its general prominence in the corpus is related to its value as a practical resource for meeting phonic requirements of rhyme. In three of the cases with proper names, the kenning construction is in an even line. This requires the determinant to carry aðalhending rhyme – i.e. the vowel must participate in the rhyme-pattern. One of these is an example with Gunnr (1a.xvii), which can be considered to engage a more specific variation of rhyme in -nn-. Skǫgul (1a.xviii) is the only determinant (or base-word) used in battle-kennings listed by Eysteinn Björnsson (–2001) that can carry rhyme in -ǫgl-. In the line above, Skǫgul rhymes with gǫgl = ‘goslings’ (sg. gagl). The only other case in which I have found gǫgl carrying aðalhending rhyme, it also rhymes with Skǫgul (Gizsv Frag 1III.4; cf. Þmáhl Máv 10V.1). Moreover, Skǫgul is the only determinant of those listed which can carry rhyme in -gl-, while the only base-word with this capacity is hagl = ‘hail’.45 This produces a high probability that Skǫgul has a functional role as a resource in composition related to particular phonic demands. The Odin-heiti Þróttr (4a.ii) is similarly the only determinant listed capable of carrying rhyme in -ótt-, which is only otherwise attested for battle kennings in the exceptional base-word íþrótt = ‘sport’ (cf. also -ót- only found in the determinant spjót = ‘lance’, while also in 45 Although not exhibiting exclusivity, hagl carries aðalhending rhyme with gagl in ESk Ingdr 3II.6; Hfr Hákdr 3III.2; SnSt Ht 62III.2; Tindr Hákdr 3I.4. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 130 Frog the base-word mót = ‘meeting’ and a few others that are less frequent). Skǫgul and Þróttr are terms associated with mythology found as determinants in several battle-kennings. Their realizations of 12(p)XYy battle-kennings can be directly associated with the fulfilment of metrical requirements for which other heiti capable of completing the kenning are lacking. Gǫndul (3b.vi) is the only determinant listed by Eysteinn Björnsson (–2001) capable of aðalhending rhyme in -ǫnd-, but in the present case appears carrying skothending rhyme in the much more common rhyme of -nd-. The heiti real- izing attested battle-kennings show that the poetic register was well equipped for this rhyme, with more frequent determinants of brandr = ‘brand’ = ‘sword’ (cf. 2a.i, 4b.i–ii), randr = ‘edge’ = ‘shield’ (cf. 2a.xiv, 3b.ii–iii), Þundr (Odin) as well as Hrund (valkyrie) and some more peripheral terms, not to mention base-words such as vindr = ‘wind’ and fundr = ‘meeting’. However, Gǫndul is the only such determinant that can also carry g-alliteration as in the example above. The ambiguity of this case can be cross-referenced laterally with other uses of Gǫndul in the poetry. A rapid survey of the Skaldic Database (which cannot be considered exhaustive at this time) reveals that, across all meters of skaldic verse, the genitive inflection Gǫndlar invariably appears carrying alliteration or (in a few cases) carrying aðalhending rhyme. This suggests that it has a functional role in composition associated with alliteration as well as with a rhyme which it is unusually equipped to carry. Correlating this obser- vation with the examples of Gunnr, Skǫgul and Þróttr, it is reasonable to view use of Gǫndul in this 12(p)XYy battle-kenning construction as conditioned by phonic demands of the meter. The interface of determinants in kennings realizing basic type 12(p)XYy suggests that word-choice has been conditioned by metrical demands, but it does not necessarily follow that the determinant is selected on the basis of a preceding rhyme-word rather than vice versa. The whole formula or even the basic type itself may also have been selected in a more reciprocal dia- lectic between form and sense in the production of the particular line. The valkyrie-name Gunnr is striking for its marked predominance in basic type 12(p)XYy rather than varying with the valkyrie-name Gǫndul. Gǫndul is the only other name of a valkyrie or of Odin in attested battle-kennings which could carry g-alliteration as a determinant in this battle-kenning basic type (following Björnsson –2001). Gǫndul appears in general to be more frequent in genitive kenning constructions (Gǫndlar NP1; cf. the case studies in Parts II–III) whereas Gunn- is common in forming compounds (Gunn-NP1). Use in 12(p XYy contradicts this pattern. The appearance of Gunnr rather than Gǫndul may therefore, at least potentially, not be independent of its clearly conventionalized use with gnýr. In other words, the use of type 12(p)XYy 131 in odd lines producing g-alliteration may have been highly conventionalized rather than freely generative. To the degree that this formula had crystallized at the lexical level, it would be resistant to alternation between Gunnr and Gǫndul as determinants. The valkyrie-name Gǫndul rather than Gunnr appears in the only odd-line example of this battle-kenning basic-type in which g-alliteration is not carried by gnýr. Use of the base-word glymr = ‘crash’ is not conditioned by rhyme-constraints and gnýr would be appropriate in this position. This makes it interesting to observe the possibility that co-variation in both the determinant and base-word is somehow related to deviation from the other- wise highly conventionalized gnýr Gunnar kenning. It becomes more note- worthy because the determinant carries the rhyme in the line. If the absence of Gǫndul from evidence of 12(p)-gný-Yy is not accidental, this suggests that use of this gnýr-formula in odd lines was metrically entangled with the rhyme in the line, and the use of an alternative rhyme with g-alliteration in the basic type could then motivate variation in the base-word. This review of vocabulary associated with mythology in battle-kennings of the 12(p)XYy metric-structural type does not show that heiti were never strate- gically employed for meaning-generation, nor has that been the purpose here. The survey of use of genitive Gǫndlar indicates a direct relationship between uses of this term and phonic demands of the meter (in contrast to, e.g., dynr). When a single example is observed in isolation, participation in alliteration and rhyme can be interpreted as a strategy to emphasize the name and the valkyrie as a mythic being. Focus here is, however, on considering such names as inte- grated resources of the lexicon. Turning attention from the use of these names in kennings generally to their contextualized occurrence in metrically entangled kennings as semantic formulae allows their use to be situated in the context of lexical variation of the particular basic type. Type 12(p)XYy is exceptional in the degree to which it displays crystallization in the use of determinants and, in odd lines, also of kennings as crystallized alliterating expressions. The kenning Gunnar gnýr as a crystallized formula accounts for 5 of the 9 uses of mythic names in this basic type. If Gunnar dynr is seen as an even-line variation on that crystallized formula, this accounts for fully two thirds of these examples. Personal names do not otherwise exhibit prominence in battle-kennings of this basic type. However, situating these names within the context of varia- tion of this type highlights the conventionalized usage of the kenning Gunnar gnýr on the one hand, while simultaneously foregrounding the fact that other mythic names in this basic type appear interfaced with phonic demands of the meter and appear to function like other determinants in battle-kennings. Turning emphasis from the names themselves to the metrically situated ken- nings illustrates the degree to which these terms are integrated into variation Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 132 Frog with other determinants of different semantic categories such as sverðr = ‘sword’ and hjalmr = ‘helmet’. This was especially evident in the variation producing the complex kenning Gunnar gagls gnýr, which shifted the determinant to a different semantic equivalence class without affecting the form of the oth- erwise crystallized formula. Consequently, when considering the use of the conventionalized kennings gnýr geira and gnýr Gunnar, in many if not most cases, the only difference between a ‘the sound of a valkyrie’ and ‘the sound of a spear’ was a question of which had the better ring to it.46 References Abram, Christopher 2011. Myths of the Pagan North: The Gods of the Norsemen. London: Continuum. Acker, Paul 1998. Revising Oral Theory: Formulaic Composition in Old English and Old Icelandic Verse. New York: Garland. Aðalsteinsson, Ragnar Ingi 2005. Nokkrar athuganir á hlutföllum orðflokka í stuðlasetningu. In: Íslenskt mál og almenn málfræði 27, 171–188. Amory, Frederic 1982. Towards a Grammatical Classification of Kennings as Compounds. In: Arkiv för Nordisk Filologi 97, 67–80. Árnason, Kristján 1991. The Rhythms of dróttkvætt and Other Old Icelandic Metres. Reykjavík: University of Iceland, Institute of Linguistics. Bell, Catherine 1992. Ritual Theory, Ritual Practice. Oxford: Oxford University Press. Birgisson, Bergsveinn 2007. Inn i skaldens sinn: Kognitive, estetiske og historiske skatter i den norrøne skaldediktingen. PhD dissertation, University of Bergen. https://bora.uib.no/bitstream/handle/1956/2732/Dr.Avh._B_Birgisson. pdf ?sequence=1 (accessed February 24, 2013). Björnsson, Eysteinn –2001. Lexicon of Kennings and Similar Poetic Circumlocutions: The Domain of Battle. https://notendur.hi.is/~eybjorn/ugm/kennings/kennings.html (accessed July 24, 2013). Blackburn, Stuart H. 1988. Singing of Birth and Death: Texts in Performance. Philadelphia: University of Pennsylvania Press. 46 I would like to thank my two anonymous reviewers for comments and observations that helped to strengthen this paper. Research presented here has been accomplished within the framework of the Academy of Finland Project “Oral Poetry, Mythic Knowledge and Vernacular Imagination” of Folklore Studies, University of Helsinki. https://bora.uib.no/bitstream/handle/1956/2732/Dr.Avh._B_Birgisson.pdf?sequence=1 https://bora.uib.no/bitstream/handle/1956/2732/Dr.Avh._B_Birgisson.pdf?sequence=1 133 Clunies Ross, Margaret (ed.) 2007. Poetry on Christian Subjects (2 vols.). Skaldic Poetry of the Scandinavian Middle Ages 7. Turnhout: Brepols. Clunies Ross, Margaret 2005. A History of Old Norse Poetry and Poetics. Cambridge: D. S. Brewer. Clunies Ross, Margaret; Gade, Kari Ellen; Marold, Edith; Nordal, Guðrún; Whaley, Diana; Wills, Tarrin 2012. General Introduction. In: Whaley, Diana (ed.) 2012. Poetry from the Kings’ Sagas I: From Mythical Times to c. 1035 (2. vols). Skaldic Poetry of the Scandinavian Middle Ages 1. Turnhout: Brepols, xiii–xciii. Coleman, Robert G. G. 1999. Poetic Diction, Poetic Discourse, and the Poetic Register. In: Adams, James Noel; Mayer, Roland (eds.), Aspects of the Language of Latin Poetry. Oxford: Oxford University Press. Falk, Hjalmar 1924. Odensheiti. Videnskapsselskapets Skrifter II: Hist.-filos. Klasse 10. Kristiania: Brøggers. Fidjestøl, Bjarne 1997. The Kenning System: An Attempt at a Linguistic Analysis. In: Haugen, Odd Einar; Mundal, Else (eds.), Selected Papers, trans. Foote, Peter. Viking Collection 9. Odense: Odense University Press, 16–67. Finnegan, Ruth 1977. Oral Poetry: Its Nature, Significance and Social Context. Cambridge: Cambridge University Press. Foley, John Miles. 1976. Formula and Theme in Old English Poetry. In: Stolz, Benjamin A.; Shannon, Richard S. (eds.), Oral Literature and the Formula. Ann Arbor: Center for the Coördination of Ancient and Modern Studies, University of Michigan, 207–232. Foley, John Miles 1988. The Theory of Oral Composition: History and Methodology. Bloomington: Indiana University Press. Foley, John Miles 1991. Traditional Oral Epic: The Odyssey, Beowulf, and the Serbo- Croation Return Song. Los Angeles: University of California Press. Foley, John Miles 1995. The Singer of Tales in Performance. Bloomington: Indiana University Press. Foley, John Miles 1999. Homer’s Traditional Art. University Park: Pennsylvania State University Press. Foley, John Miles 2002. How to Read an Oral Poem. Urbana: University of Illinois Press. Frank, Roberta 1978. Old Norse Court Poetry: The Dróttkvætt Stanza. Islandica 42. Ithaca: Cornell University Library. Frank, Roberta 1985. Skaldic Poetry. In: Clover, Carol J.; Lindow, John (eds.), Old Norse-Icelandic Literature: A Critical Guide. Islandica 45. Ithaca: Cornell University Library, 157–196. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 134 Frog Frog 2009. Speech-Acts in Skaldic Verse: Genre, Formula and Improvisation. In: Dewey, Tonya Kim; Frog (eds.), Versatility in Versification: Multidisciplinary Approaches to Metrics. Berkeley Insights in Linguistics and Semiotics 74. New York: Peter Lang, 223–246. Frog 2010. Listening to Poetry in a Dead Language: Could dróttkvætt Have Been Trochaic? In: RMN Newsletter [1], 23–28. Frog 2011a. Alvíssmál and Orality I: Formula, Alliteration and Categories of Mythic Being. In: Arkiv för Nordisk Filologi 126, 17–71. Frog 2011b. Distinguishing Continuities. Textual Entities, Extra-Textual Entities and Conceptual Schemas. In: RMN Newsletter 2, 7–15. Frog 2012. On the Case of Vambarljóð II: Register and Mode from Skaldic Verse to sagnakvæði. In: RMN Newsletter 5, 49–61. Frog 2013. The (De)Construction of Mythic Ethnography I: Is Every þurs in Verse a þurs? In: RMN Newsletter 6, 52–72. Frog (forthcoming). Metrical Entanglement and Grammaticization: Kennings, Constructions and Compositional Strategies in Skaldic Verse. In: Árnason, Kristján; Aðalsteinsson, Ragnar Ingi; Eyþórsson, Þórhallur; Carey, Stephen Mark (eds.), Approaches to Germanic Metre [working title]. Reykjavík: University of Iceland Press. Frog; Roper, Jonathan 2011. The Vanir versus Verses: Response to Simek’s ‘Vanir Obituary’. In: RMN Newsletter 2, 29–37. Frog; Stepanova, Eila 2011. Alliteration in (Balto-)Finnic Languages. In Roper, Jonathan (ed.), Alliteration in Culture. Houndmills: Palgrave MacMillan, 195–218. Gade, Kari Ellen 1994. On the Recitation of Old Norse Skaldic Poetry. In: Uecker, Heiko (ed.), Studien zum Altgermanischen: Festschrift für Heinrich Beck. Ergänzungsbände zum Reallexikon der germanischen Altertümskunde 11. Berlin: de Gruyter, 126–151. Gade, Kari Ellen 1995. The Structure of Old Norse dróttkvætt Poetry. Ithaca: Cornell University Library. Gade, Kari Ellen (ed.) 2009. Poetry from the Kings’ Sagas II: From c.1035 to c.1300 (2 vols.). Skaldic poetry of the Scandinavian Middle Ages 2. Turnhout: Brepols. Gil’ferding, A. F. 1894. Onezhskija byliny, zapisannyja Aleksandrom Fedorovichem Gil’ferdingom letom 1871 goda I. St. Petersburg. Goldberg, Adele 2006. Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press. Grove, Jonathan 2008. Skaldic Verse-Making in Thirteenth-Century Iceland: The Case of the Sauðafellsferðarvísur. In: Viking and Medieval Scandinavia 4, 85–131. Hainsworth, John B. 1968. The Flexibility of the Homeric Formula. Oxford: Clarendon Press. 135 Halliday, Michael A. K. 1978. Language as Social Semiotic: The Social Interpretation of Language and Meaning. London: Edward Arnold. Harris, Joseph 1983 [2008]. Eddic Poetry as Oral Poetry: The Evidence of Parallel Passages in the Helgi Poems for Questions of Composition and Performance. In: Harris, Joseph, “Speak Useful Words or Say Nothing”: Old Norse Studies. Descis, S. E.; Hill, T. D. (eds.). Islandica 53. Ithaca: Cornell University Library, 189–225. Harvilahti, Lauri 1992a. Kertovan runon keinot: Inkeriläisen runoepiikan tuottamisesta. Helsinki: Suomalaisen Kirjallisuuden Seura. Harvilahti, Lauri 1992b. The Production of Finnish Epic Poetry: Fixed Wholes or Creative Compositions? In: Oral Tradition 7, 87–101. Harvilahti, Lauri 2000. Variation and Memory. In: Honko, Lauri (ed.), Thick Corpus, Organic Variation and Textuality in Oral Tradition. Studia Fennica Folkloristica 7. Helsinki: Finnish Literature Society, 57–75. Harvilahti, Lauri 2003. The Holy Mountain. FF Communications 282. Helsinki: Academia Scientiarum Fennica. Haymes, Edward R. 2004. The Germanic Heldenlied and the Poetic Edda: Speculations on Preliterary History. In: Oral Tradition 19, 43–62. Holoka, James P. 1976. The Oral Formula and Anglo-Saxon Elegy: Some Misgivings. In: Neophilologus 60, 570–576. Holland, Gary 2005. Kennings, Metaphors, and Semantic Formulae in Norse dróttkvætt. In: Arkiv för Nordisk Filologi 120, 123–147. Hollmérus, Ragnar 1992. Studier över allitterationen i Eddan. Helsinki: Svenska litteratursällskapet i Finland. Honko, Lauri 1998. Textualising the Siri Epic. FF Communications 264. Helsinki: Academia Scientiarum Fennica. Jónsson, Finnur (ed.) 1967. Den Norsk-Islandske Skjaldedigtning A–B. København: Rosenkilde og Bagger. Kellogg, Robert 1965. The South Germanic Oral Tradition. In: Bessinger, Jess B.; Creed, Robert P. (eds.), Franciplegius: Medievaland Linguistic Studies in Honor of Francis Peabody Magoun, Jr. New York: New York University Press, 66–74. Kiparsky, Paul 1976. Oral Poetry: Some Linguistic and Typological Considerations. In: Stolz, Benjamin A.; Shannon, Richard S. (eds.), Oral Literature and the Formula. Ann Arbor: Center for the Coördination of Ancient and Modern Studies, University of Michigan, 73–106. Kuhn, Hans 1983. Das Dróttkvætt. Heidelberg: Carl Winter. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? 136 Frog Kuusi, Matti 1975. ‘The Bridge and the Church’: An Anti-Church Legend. In: Leino, Pentti; Kaivola-Bregenhøj, Annikki; Vento, Urpo (eds.), Finnish Folkloristics 2. Studia Fennica 18. Helsinki: Finnish Literature Society, 37–75. Leino, Pentti 1970. Struukturaalinen alkusointu suomessa: Folklorepohjainen tilastoanalyysi. Helsinki: Suomalaisen Kirjallisuuden Seura. Lindow, John 1975. Riddles, Kennings, and the Complexity of Skaldic Poetry. In: Scandinavian Studies 47(3), 311–327. Lindow, John 1985. Mythology and Mythography. In: Clover, Carol J.; Lindow, John (eds.), Old Norse-Icelandic Literature: A Critical Guide. Islandica 45. London: Cornell University Press, 21–67. Lindow, John 2001. Norse Mythology: A Guide to Gods, Heroes, Rituals, and Beliefs. Oxford: Oxford University Press. Lönnroth, Lars 1971. Hjalmar’s Death-Song and the Delivery of Eddic Poetry. In: Speculum 46, 1–20. Lönnroth, Lars 2002. The Founding of Miðgarðr (Vǫluspá 1–8). In: Acker, Paul; Larrington, Carolyn (eds.), The Poetic Edda: Essays on Norse Mythology. New York: Routledge, 1–25. Lord, Albert Bates 1960. The Singer of Tales. Harvard Studies in Comparative Literature 24. Cambridge: Harvard University Press. Lord, Albert Bates 1991. Epic Singers and Oral Tradition. Ithaca: Cornell University Press. Lord, Albert Bates 1995. The Singer Resumes the Tale. Lord, Mary Louise (ed.). London: Cornell University Press. Marold, Edith 1983. Kenningkunst: Ein Beitrag zu einer Poetik der Skaldendichtung. Quellen und Forschungen zur Sprach- und Kulturgeschichte der germanischen Völker, Neue Folge 80. Berlin: de Gruyter. Meissner, Rudolf 1921. Die Kenningar der Skalden: Ein Beitrag zur skaldischen Poetik. Bonn: Schroeder Mellor, Scott 1999 [2008]. Analyzing Ten Poems from The Poetic Edda: Oral Formula and Mythic Patterns. Mellen: Lewiston. (Publication of Function and Formula: An Analysis of Ten Poems from the Codex Regius. PhD dissertation, University of Wisconsin, Madison, 1999.) Noreen, Erik 1921. Studier i fornvästnordisk diktning. Uppsala Universitets Årsskrift 1921, Filosofi, Språkvetenskap och Historiska Vetenskaper 4. Uppsala: A.B. Akademiska Bokhandeln. Osborne, Emily Margaret 2012. Thinking Outside the Hall: The Conceptual Boundaries of Skaldic Verse. Unpublished PhD dissertation, University of Cambridge. 137 Osborne, Emily 2013. Categorising Christ within an Age-ǫld Paradigm: The ‘Kenning System’ and Shifting Cultural Referents. In: Frog; Latvala, Pauliina; Leslie, Helen F. (eds.), Approaching Methodology (2nd rev. edn.). Annales Academiae Scientiarum Fennicae Humaniora 368. Helsinki: Academia Scientiarum Fennica, 331–346. Parry, Milman 1928. L’épithète traditionnelle dans Homère. Paris: Société d’Éditions “Les Belles Lettres”. Parry, Milman 1930. Studies in the Epic Technique of Oral Verse-Making. I. Homer and Homeric Style. In: Harvard Studies in Classical Philology 41, 73–147. Perkins, Richard 1982–1985. Rowing Chants and the Origins of Dróttkvæðr Háttr. In: Saga-Book of the Viking Society 21, 155–221. Potts, Deborah 2011. A Cognitive Approach to the Analysis of the Extant Corpus of Kennings for Poetry. Unpublished PhD dissertation, University of Cambridge. Price, Neil S. 2002. The Viking Way: Religion and War in Late Iron Age Scandinavia. Uppsala: Department of Archaeology and Ancient History. Roper, Jonathan 2012. Synonymy and Rank in Alliterative Poetry. In: Sign Systems Studies 40(1–2), 82–93. Rubin, David C. 1995. Memory in Oral Traditions: The Cognitive Psychology of Epic, Ballads, and Counting-Out Rhymes. Oxford: Oxford University Press. Saarinen, Jukka 2013. Behind the Text: Reconstructing the Voice of a Singer. In: RMN Newsletter 7, 34–43. Salminen, Väinö 1934. Suomalaisten muinaisrunojen historia. Helsinki: Suomalaisen Kirjallisuuden Seura. Schmitt, Norbert; Carter, Ronald 2004. Formulaic Sequences in Action: An Introduction. In: Schmitt, Norbert (ed.), Formulaic Sequences: Acquisition, Processing and Use. Amsterdam: John Benjamins. Sievers, Eduard 1893. Altgermanische Metrik. Sammlung kurzer Grammatiken germanischer Dialekte Erzänzungsreihe 2. Halle: Niemeyer. Siikala, Anna-Leena 1990 [1984]. Interpreting Oral Narrative. FF Communications 245. Helsinki: Academia Scientiarum Fennica. Skaldic Database. http://abdn.ac.uk/skaldic/db.php (accessed December 4, 2013). Stepanova, Eila 2011. Reflections of Belief Systems in Karelian and Lithuanian Laments: Shared Systems of Traditional Referentiality? In: Archaeologia Baltica 15, 128–143. Stepanova, Eila 2012. Mythic Elements of Karelian Laments: The Case of syndyzet and spuassuzet. In: Frog; Siikala, Anna-Leena; Stepanova, Eila (eds.), Mythic Discourses: Studies in Uralic Traditions. Studia Fennica Folkloristica 20. Helsinki: Finnish Literature Society, 257–287. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? http://abdn.ac.uk/skaldic/db.php 138 Frog Stepanova, Eila forthcoming. Karjalan Seesjärven itkukieli. PhD thesis, University of Helsinki. Sullivan, Karen 2008. Genre-Dependent Metonymy in Norse Skaldic Poetry. In: Language and Literature 17(1), 21–36. Sverdlov, Ilya V. 2003. Extra Elements in Kennings, or ‘Different Ways to Feed the Wolf Single-Wordedly’. In Skandinavistik 33(2), 101–113. Sverdlov, Ilya V. 2006. Kenning Morphology: Towards a Formal Definition of the Skaldic Kenning, or Kennings and Adjectives. In: McKinnell, John; Ashurst, David; Kick, Donata (eds.), The Fantastic in Old Norse / Icelandic Literature – Sagas and the British Isles. Preprint papers of the 13th International Saga Conference, Durham and York 6–12 August 2006. Durham: The Centre for Medieval and Renaissance Studies, Durham University, 928–934. Sykäri, Venla 2011. Words as Events: Cretan Mantinádes in Performance and Composition. Studia Fennica Folkloristica 18. Helsinki: Finnish Literature Society. Tarkka, Lotte 2005. Rajarahvaan laulu: Tutkimus Vuokkiniemen kalevalamittaisesta runokulttuurista 1821–1921. Helsinki: Suomalaisen Kirjallisuuden Seura. Þorgeirsson, Haukur 2010. Gullkársljóð og Hrafnagaldur: Framlag til sögu fornyrðislags. In: Gripla 21, 229–334. Þorgeirsson, Haukur 2013. Poetic Formulas in Late Medieval Icelandic Folk Poetry: The Case of Vambarljóð. In: Frog; Latvala, Pauliina; Leslie, Helen F. (eds.), Approaching Methodology (2nd rev. edn.). Annales Academiae Scientiarum Fennicae Humaniora 368. Helsinki: Academia Scientiarum Fennica, 347–375. Thorvaldsen, Bernt Øyvind 2006. Svá er sagt í fornum vísindum: Tekstualiseringen av de Mytologiske Eddadikt. PhD dissertation University of Bergen. https:// bora.uib.no/bitstream/handle/1956/2236/Dr_Avh_Bernt_Oyvind_Thorvaldsen. pdf ?sequence=1 (accessed January 1, 2011). Thorvaldsen, Bernt Øyvind 2008. Om Þrymskviða, tekstlån og tradisjon. In: Maal og Minne 2008, 142–166. Turville-Petre, Edward Oswald Gabriel 1964. Myth and Religion in the North: The Religion of Ancient Scandinavia. New York: Holt, Rinehart & Winston. Turville-Petre, Edward Oswald Gabriel 1976. Scaldic Poetry. Oxford: Clarendon Press. de Vries, Jan 1934. De Skaldenkenningen met mythologischen Inhoud. Nederlandsche Bijdragen op het Gebied van germaansche Philologie en Linguistiek 4. Haarlem: H.D. Tjeenk Willink & Zoon N.V. de Vries, Jan 1956–1957. Altgermanische Religionsgeschichte I–II. 2nd edn. Berlin: de Gruyter. https://bora.uib.no/bitstream/handle/1956/2236/Dr_Avh_Bernt_Oyvind_Thorvaldsen.pdf?sequence=1 https://bora.uib.no/bitstream/handle/1956/2236/Dr_Avh_Bernt_Oyvind_Thorvaldsen.pdf?sequence=1 https://bora.uib.no/bitstream/handle/1956/2236/Dr_Avh_Bernt_Oyvind_Thorvaldsen.pdf?sequence=1 139 Whaley, Diana 2005. Skaldic Poetry. In: Mcturk, Rory (ed.), A Companion to Old Norse-Icelandic Literature and Culture. London: Blackwell, 479–502. Wills, Tarrin 2009. The Development of Skaldic Language. In: Ney, Agneta; Williams, Henrik; Ljungqvist, Fredrik Charpentier (eds.), Á austrvega: Saga and East Scandinavia. Preprint Papers of the 14th International Saga Conference, Uppsala, 9th–15th August 2009. Gävle: University of Gävle, 1032–1038. Wills, Tarrin n.d. a. Analysis: Features by Lemmatised Line (Bars) (Various Features by Line of Lemmatised helmingr). https://www.abdn.ac.uk/skaldic/db.php?id=1 40&if=default&table=php_reports&val=&view= (accessed December 12, 2013). Wills, Tarrin n.d. b. Analysis: Features by Line (Various Features by Line of dróttkvætt helmingr). https://www.abdn.ac.uk/skaldic/db.php?id=135&if=default&table=p hp_reports&val=&view= (accessed December 12, 2013). Wray, Alison 2002. Formulaic Language and the Lexicon. Cambridge: Cambridge University Press. Würth, Stephanie 2007. Skaldic Poetry and Performance. In: Quinn, Judy; Heslop, Kate; Wills, Tarrin (eds.), Learning and Understanding in the Old Norse World: Essays in Honour of Margaret Clunies Ross. Medieval Texts and Cultures of Northern Europe 18. Turnhout: Brepols, 263–281. Mythological Names and dróttkvætt Formulae I: When is a Valkyrie Like a Spear? https://www.abdn.ac.uk/skaldic/db.php?id=140&if=default&table=php_reports&val=&view= https://www.abdn.ac.uk/skaldic/db.php?id=140&if=default&table=php_reports&val=&view= https://www.abdn.ac.uk/skaldic/db.php?id=135&if=default&table=php_reports&val=&view= https://www.abdn.ac.uk/skaldic/db.php?id=135&if=default&table=php_reports&val=&view=