Studia Metrica et Poetica sisu 4_1.indd Prototypes and structures in eddic poetry Seiichi Suzuki, The Meters of Old Norse Eddic Poetry: Common Germanic Inheritance and North Germanic Innovation (Ergänzungsbände zum Reallexikon der Germanischen Altertumskunde, Band 86). Berlin: Walter de Gruyter, 2014. XLV+1096 pp. Kristján Árnason*1 Introduction The volume entitled The Meters of Old Norse Eddic Poetry (hence MONEP) is an impressive book, more than 1100 pages in all, and it can truly be called a landmark in the field of Nordic poetics. It is the latest contribution to a series of publications by the same author in the field of Germanic metrics. In a series of publications (e. g. Suzuki 1995, 1996, 2004, 2008, 2009, 2010, 2011, 2014, 2016) Suzuki has developed a model for how different Germanic cultures (Old English, German and Nordic) developed different styles and metrical patterns, each creating, so to speak, its own variant or variants of the common metrical proto-form. The most important of these works are the volumes on Old English Beowulf (1996) and Old Saxon poetry (2004), and MONEP closes the cycle. The model carries on the Sievers tradition, but it provides a new and, in some respects at least, more theoretically sophisticated interpretation. Sievers’ five type system Sievers’ five-type model has practically dominated the field of Old Germanic metre since late nineteenth century, not least in English philological schol- arship (cf. Bliss 1967, Fulk 1992; for a recent presentation of the model see Terasawa 2011). In spite of some attempts at replacing the model, such as Russom’s word foot theory (Russom 1998, 2009) or Cable’s theory of positions and strength relations (Cable 1974: 84 ff.), the five type model is still going * Author’s address: Kristján Árnason, The University of Iceland, Árnagarði v. Suðurgötu, 101 Reykjavík, Iceland. E-mail address: kristarn@hi.is. Studia Metrica et Poetica 4.1, 2017, 104–128 doi: doi.org/10.12697/smp.2017.4.1.05 105Prototypes and structures in eddic poetry strong. The later models and revisions are designed to describe the metre of Beowulf, but mutatis mutandis they should also be applicable to German and Icelandic. Still, in a recent handbook of eddic poetry (Larrington et al. 2016), Fulk’s description of the eddic metres basically follows Sievers, hardly men- tioning other approaches or insights. This shows the strength of the paradigm, at least in some circles. The original five type system was intended to improve earlier theories about the structure of traditional Germanic poetry, basically what was called die Zweihebungstheorie, or with respect to the long line: Vierhebungstheorie. This two or four lift model had been developed by scholars like Lachmann and Wackernagel, but Sievers pointed out that the putative “regellosigkeit in der behandlung des auftakts und der senkungen” (‘the unsystematic treatment of the anacrusis (upbeat) and the low’, Sievers 1893: 8) implied by the simple four lift model is not what the text shows. Sievers maintains that the form of the drop is not as free as the model would predict. According to Sievers, each short line had four glieder (or positions), two lifts and two drops in addition to an optional upbeat (auftakt), and certain principles prevail regarding their form and distribution (cf. Sievers 1893: 29 ff.). The lifts and drops combine to form feet of different shapes and sizes; the relation between the composition of lifts and drops was not free, nor the relation between the two feet within the line. And in a way the basic unit of analysis is the half-line (verse, first: a-verse and second: b-verse), which can be realized in different ways according to the five-type system. It is clear, however, that the five-type system was not meant to reject the earlier insights, and in fact Sievers seems hesitant to call his analysis a theory. Before this could be done, he says: “[müsste mann] in der durchgeführten einordnung gewisser vielleicht mehrdeutiger einzelformen in bestimmte typenschemata etwas besonders theoretisches finden wollen”1 (Sievers 1893: 8–9, italics mine). And he calls for insights about musical characteristics and recitation or performance of the text. The type system is then basically a descriptive tool, and not a proper theory. Also, Bliss (1967: 106 ff.) acknowl- edges the limits of the typology and the statistics. The weaknesses of the theoretical foundations of Sievers’ system, as pointed out many times (cf. e. g. Russom 1998, Yakovlev 2008, Goering 2016), are several. To take an example, although in the original model the verses (or half-lines) were divided into two feet, there seems to be some sort of “global” 1 ‘One would have to find something specifically theoretical behind the actual arrangements of certain perhaps ambiguous forms into specific type-schemata’. 106 Kristján Árnason view on the line, so that lack of material in one part of the line may be com- pensated for by extra material in the other. This is the case in types D and E; in Sievers’ words: “[W]eil dem einen Fusse die Senkung fehlt, der andere dafür eine zweigliedrige Senkung, genauer gesagt eine Nebenhebung und eine eigentliche Senkung besitzt”2 (1885: 3). This global view on the relation between the metrical constituents means that the shape of the whole line is a structural parameter, which raises the question of the foot as an independently motivated unit of structure (cf. e. g. Suzuki 1996: 35–44, where it is concluded that the foot is not a significant part of the structure). What are the limits on the combination of feet within lines: since we have trisyllabic feet like / \ x and / x \, why do we not have lines like / \ x | / x \ , and so on? These questions are not answered (and should not be, according to Sievers, if we follow the interpretation that the system is only descriptive). On top of this, many of the lines that occur, e. g. in eddic fornyrðislag, do not actually fit any of the basic five type patterns. Thus, as shown in Árnason (2016: 93), there are 138 (2.1%) trisyllabic lines in the fornyrðislag poems in the Codex Regius of the Poetic Edda. To account for these lines, as well as lines which have only two positions, new types would have to be assumed (F and G, cf. Sievers 1893: 67–68), clearly breaking the five-type frame. Another similar fact, not covered, is the relation between the two half-lines, forming the long line. This is not quite symmetrical, since generally the on-line (a-verse) shows more freedom in composition than the off-line (b-verse) (cf. Sievers 1893: 24). It is not clear exactly how these “global” relations or compensation between feet and lines is governed. Thus, it seems that, for Sievers, the type model was basically a taxonomic tool, describing the text, whereas the ultimate aim of the metrical analysis, what he calls (op. cit., p. 24) vollständige rhythmisierung, which would involve an account of scansion and recitation (vortrag), should be based, in addition to the text itself, on insights from history or the evolution of the forms, and the history of music (entwicklungs- und musikgeschichtliche gesichtspunkte). The work presented in his book from 1893 was clearly not intended to reach this goal, as shown by the following quote: “Die folgenden darstellungen beschränken sich … absichtlich auf die positiven bestimmungen, die sich aus der untersuchung des sprachlichen substrats der verse (des rhythmizómenon) ergeben”3 (op. cit., p. 24). That is, the intention is to give a detailed description 2 ‘Since one foot lacks a low, therefore the other has a double low or more precisely a second- ary lift and a proper low’. 3 ‘The following presentation is intentionally limited to the positive conclusions which can 107Prototypes and structures in eddic poetry of the rhythmic or “rhythmisized” text, or what might be called composition, rather than to give a full account of the form and the actual performance (vortrag) (cf. Allen 1973: 105, Árnason 2000 [1991]: 4). And already early scholars like Henry Sweet (cf. Bliss 1967: 3) and Andreas Heusler (1969 [1889]: 693) agree with Sievers: the five-type system is not a theory, but a typical pre- theoretical taxonomy. The description of variation within the normal metrical forms given by Sievers (1893: 25–31 for Germanic verse in general, and 65–66 ff. for Nordic verse) is often far from clear or explicit. To mention a case of such confusion: in the mapping between metrical form and linguistic form, the main principle is that lifts (hebungen) are represented (vertreten) by a heavy (or long) sylla- ble. But there are at least two types of exceptions from this. One is resolution (auflösung, cf. op. cit., p. 27). According to this principle the lift is represented by a short syllable followed by an unstressed syllable. This is made possible in performance by some sort of compression of the two syllables, named ver- schleifung. Other instances of “shortening of the lift” (verkürzung der hebung), which occur in Nordic poetry, are listed later (op. cit., p. 65). One is when the first syllable of words like konungar ‘kings’, a short first syllable followed by a heavy one, is allowed to form a lift. Here the law of quantity is broken in order to “preserve the secondary stress” (das quantitätsgesetz [wird] vernachlässigt, um den nebenton wehren zu können). Another instance is when secondarily stressed light syllables of words like krǫpturligan ‘powerful-ACC’ are allowed to form lifts. This is said to be due to a tendency to avoid resolution in the second lift of a line. Sievers subsumes these “licenses” under the heading of a shortening of the lift (verkürzung der hebung, p. 65), but there seems to be a fundamental differ- ence between, on the one hand, the formation of the lift by two syllables, and its representation by a single light syllable (which has been called “suspension of resolution”). The term auflösung (translated into English as ‘resolution’) implies that the lift is not weaker or shortened, since it is filled by two syllables (Sievers does not refer to moras). On the other hand, lifts formed by short (or light) syllables, seem to be weaker or “shorter” than normal ones. It is not clear how this should be interpreted. We do not know whether this is a special type of lift (defined by some metrical rule or licence), or whether the lift is made shorter (textually or in performance) by the use of the light syllable. Adding to the confusion, secondary lifts (nebenhebungen) are assumed for trisyllabic be drawn from the linguistic substrate of the verse (the rhythmizómenon)’, cf. also Heusler 1969 [1889]: 693, where this point is emphasized. 108 Kristján Árnason feet. And still other exceptions allow for relatively strong drops (nebentoninge senkungen), beside the variation in the number of syllables in weak positions. These complications in the mapping relations have in later scholarship been used to produce what look (to the outsider at least) like an endless number of different realisations and subtypes. In spite of Sievers’ warning, the comments on variation in composition have been taken as deductive rules for generat- ing the different types, the modesty of the original claims seemingly being forgotten. Some scholars have of course noted the imperfections of the model, and alternative approaches have been suggested, e. g. Cable’s four-position theory (Cable 1974, Yakovlev 2008), and Russom’s word-foot theory (Russom 1998, Goering 2016). Yet, central figures in the field of Old Germanic metrics (e. g. Bliss 1967, Cable 1974 and Fulk 1992) basically interpret the system in a deductive way, it seems. A special characteristic of the type-model as it is applied in later schol- arship is that it is meant to account with the same terminology for poetry from all branches of Germanic, High German, Saxon and English, as well as Nordic. And for some scholars at least, the agenda was to reconstruct some original Germanic poetic structure, much in the same way as Germanic proto- forms have been reconstructed in order to account for the historical relations between the Germanic dialects and later languages. The similarity between the Nordic and West Germanic poetic forms are of course obvious and testify to a common heritage, e. g. regarding alliteration and the four-beat rhythm of the long line with a caesura between the a-verse and the b-verse. Yet, the different styles of e. g. Völuspá vis a vis Beowulf, Heliand and Hildbebrandslied must raise the question whether these Genres should be analysed and explained as variants of the same form. To mention just one important difference, which seems often to be overlooked or downplayed in this historical perspective, is the fact that Nordic poetry is stanzaic, i. e. dividing longer poems into basi- cally quadrilinear units, based on formal patterns (such as quadruple rhythm and alliteration), and divisions in content, giving the stanza a clear formal and functional status, whereas the West Germanic corpus is “stichic” in structure, making the line the fundamental unit of the form. The development of differ- ent eddic varieties like ljóðaháttr, fornyrðislag and málaháttr, not to mention the skaldic dróttkvætt is part of this story. 109Prototypes and structures in eddic poetry Suzuki’s work: the stochastic approach and prototypes As already emphasised, and regardless of general issues, MONEP is an impres- sive contribution to the study of Old Icelandic poetic form. Although some of the results seem to the present reviewer to be doubtful, the sheer amount of information contained in the book makes it a “must read” for any scholar studying the form and structure of Old Icelandic poetry. The book is beauti- fully laid out and written in a very clear (if lengthy) style, and its organisation makes it relatively easy to use, although obviously the sheer weight of the volume prevents it from being any kind of bedtime reading. The book is divided into three main parts, each devoted to one of the three Nordic metres, fornyrðislag, málaháttr and ljóðaháttr. The longest by far is the one on the fornyrðislag, consisting of six chapters (2–7): Chapter 2 “Verse types and their realizations”, Chapter 3 “Anacrusis and catalexis”, Chapter 4 “Resolution”, Chapter 5 “The cadence”, Chapter 6 “Alliteration”, and Chapter 7 “The stanza”. The part devoted to the málaháttr contains four chapters, Chapter 8 “The prototype of málaháttr: Atlamál in grœnlenzco”, analysing the only undisputed instance of a whole poem in the metre, and then there are three chapters on as many “peripheral variants of fornyrðislag/málaháttr”: Atlakviða (hin gænlenzka), Hamðismál and Hárbarðsljóð. The part on ljóðaháttr is three chapters, entitled “The a-verse” (Ch. 12), “The c-verse” (Ch. 13), and “The stanza” (Ch. 14). The book ends with two appendices and a four part index (scansion, authors, subjects and verses). All of this, its thoroughness in cover- ing the material and the various aspects of the poetic texts, makes MONEP a useful handbook on the subject. In his earlier publications Suzuki has applied a statistical model which shows how different Germanic cultures (English, German and Nordic) devel- oped different styles and metrical types, each creating, so to speak, its own variant or variants of the common metrical proto-form. The “Leitmotif ” in this work is an interesting theoretical twist referring to (and probably inspired by) ideas which have developed in theoretical linguistics as an alternative to the generative paradigm, known under such labels as “usage based grammar”, “prototype theory” or “exemplar theory”. In fact this forms an explanatory model which, in some sense at least, seems to legitimise the proliferation of types and subtypes, putting the data in an interesting theoretical perspective. Although based on the type system, and making use of and modifying Sievers’ classification, there are some rather substantial differences from the original model in the approach, and methodology. The idea promoted is that “meter constitutes a prototype-based, cognitive system of rules, constraints and representations, much as its linguistic 110 Kristján Árnason foundations do”, and this is “a stochastic organizing system with varying realizations, […] and subject to variation […]” (MONEP: 13–14). In this context the type system acquires a more interesting status, somewhat in the vein of modern usage based theories (Bybee 2001). The metrical prototypes develop through a sort of stylisation of linguistic forms and what might be called new or changing “poetic habits”. But at the same time Suzuki makes a clear distinction between metrical form and linguistic form. And this means that “linguistic development” and “metrical development” occur in separate spheres. The metrical system, although reflecting linguistic structure, can thus undergo change as such. Thus we learn that the Nordic form fornyrðislag became “sharply distinct” from its West Germanic cognates due to three met- rical innovations. For one thing, the realisation of the cadence by a heavy disyllable of the type kindir ‘people’ occurred “with such conspicuous fre- quency that it became closest to being an optimal cadence form” (Suzuki 2011: 378). A second Nordic innovation, according to Suzuki, was the development of catalexis, which gave lines with only three positions, which can be “derived from the normal counterpart through realization of the verse final drop by zero” (Suzuki 2011: 377, cf. Suzuki 2009). This amounts to the emergence of a new type as a metrical entity, involving a change in metrical form. A third Nordic innovation is the “generalisation of suspension of resolution whereby the string of a short syllable and an unstressed syllable of any length (px) is permitted to fill the concatenation of lift + drop in verse final position without notable restriction” (Suzuki 2011: 377). Among new subtypes or type realiza- tions are types like ljósast fyrir (Grsp 21.3), which is seen as a variant of A1 (/ x / x) with the second lift realised by a light disyllable, and Nú em ek svá fegin (HH II 43.1) xxxxpx, classified as A3- (a catalectic version of A3), with the first lift unrealised (cf. Suzuki 2014: 73; 2016). Also, “suspension of resolution” in type C: fyr mold neðan (Vsp. 2,8) gives lines with the second lift realised by a single light syllable, and Catalexis gives lines ending in two lifts: þeir er miðgarð (Vsp 4,3), which is something not allowed in the original five-type model. Thus new structures and types develop, and Suzuki systematically looks for statistical significance, which (it seems) would justify setting up a separate type or subtype. The aim is to identify significant verse types at the underlying level of metrical representa- tion, determine their manifold variations at the surface level of realization, and explore the formal and functional organization of these verse types and tokens largely on a stochastic basis. While the number and kinds of metrical positions and their linear sequencing determine a restricted set of verse types, a broad 111Prototypes and structures in eddic poetry latitude of alignment of these invariant metrical positions to diverse language materials results in a wide-ranging variation of verse types on the surface. (Suzuki 2014: 25) The aim is thus to analyse and classify the types and subtypes and to describe their distribution (between a-verse and b-verse) and the alliterative patterns (whether or not there is double alliteration in the a-verse). Interestingly enough, this statistical approach is to some extent foreshadowed in Bliss’ clas- sification and description (1967: 180–187) of the proportional distribution of types between a-verse, and b-verse and double or single alliteration in the a-verse in Beowulf. The types Chapter 2 of the book deals with the verse types and their realizations, and appropriately, the account starts with the type A1 (/ x / x), which might be considered the most prototypical of all eddic and Germanic patterns. But there is variation in the realizations of both the lifts (/) and the drops (x). Thus it may happen that the drops are ‘heavy’ or occupied by more than one syllable, and the lifts may be resolved (realized by two syllables) or light (realized by a light or short syllable by ‘suspension of resolution’). To take an example of the statistical analysis, it turns out that there is significantly more freedom in the realisation of the first drop in the a-verse compared to the b-verse, although this is not seen as motivating the postulation of new types or subtypes. The general question regarding the A-types seems to be twofold: How do we delimit the set of A-types (or rather the class of A-types), and what sig- nificance is there to be assigned to the label A? And in relation to that, we may ask how many different (proto-)types we want to have. An example of the argumentation is to be found on pp. 43–45, where a motivation is given for distinguishing a special subtype A2a ( / \ / x), where the second posi- tion is filled by a second component of a compound (e. g. Sg 61.3 frumver sínum). It turns out that this configuration (PS#Px) differs significantly from A1-lines with the configuration Px#Px (e. g. Vsp. 1.2 helgar kindir). There is thus a significant difference “both in terms of verse distinction”, i. e. in that the distribution between a-verses and b-verses is not equal, and as regards the alliterative pattern. Thus the former type (involving compounds) “fails to show the marked preference for the b-verse and single alliteration” (p. 44), which is characteristic of A1. And there is a tendency for the A2-type to have 112 Kristján Árnason double alliteration whereas the opposite is the case for the normal A1-type, which seems to prefer single alliteration. The basic reasoning is, then, that if a significantly different distribution can be found between two types, there is reason to assume different types or subtypes, such as A1 vs. A2. But an obvious question in this connection would be to ask whether there is an explanation (e. g. a structural one) for the statistical differences in dis- tribution, between A1 and A2, in other words, we might want to look for the substantive reasons for the statistical patterns. Two types of consideration come to mind. For one thing, it would seem that the preference for the most basic type (A1) to occur in the b-verse might be due to the very well-known fact that metrical regularity normally increases toward the end of constituents. The trochaic pattern of the A1 type is by far the most common verse type, and it can also be seen as the most natural one, both from the linguistic point of view (e. g. in terms of stress patterns of words, disyllabic trochaic words, e. g. involving stems and inflectional endings) and metrical one (since a Strong – Weak rhythm forms a good basis for creating rhythmic regularity in a text). Many cases of this phenomenon have been observed in the literature (cf. e. g. Hayes & McEachern 1998). The other consideration which seems to be relevant here is the function of alliteration and its relation to metrical strength. It seems to be a dictum amongst many scholars in the Sieversian tradition that there is a fixed relation between alliteration and strength or ictus, which would mean that an alliterat- ing syllable should be stronger than a corresponding or otherwise “equal” one which does not alliterate. But as I have argued (Árnason 2007) the function of alliteration (and rhyme) is not a rhythmic one to do with accentuation: like other rhyme, it serves the function of connecting parts of the text and forming constituents, in the case of eddic and dróttkvætt a pair of lines, and in the case of West Germanic, the long line. The rhythmic implications of this binding function are only indirect. It is not clear that alliteration as such implies added strength or emphasis, so that the alliterating staves necessarily have stronger stress in performance or that the lifts containing them have to be stronger than the ones that do not. Since the function of alliteration is to form a connection between a- and b-verse, it is only to be expected that more “material” in the a-line should be conducive to additional marking of this type. An interesting case when it comes to defining subtypes belonging to the A-class is the one labelled A1s by Suzuki (Sievers’ A2k). This type is character- ised as realizing the second lift by a short (light) stressed syllable. Examples are lines like Hym 4.7 ástráð mikit, Gðr II 41.7 sorgmóðs sefa, Bdr 9.2 hróðrbarm þinig (p. 33), where the first foot is realised by a compound (PS#px), and HH II 11.6 hildings synir, Grp 12.5 leið at huga Hdl 1.6 ríða við scolom (p. 36), 113Prototypes and structures in eddic poetry where the first drop is, according to Suzuki, realized by a weak syllable (i. e. with the composition Px#px). In the former case the second lift is realized by a short stressed syllable on its own “through suspension of resolution”, and this is motivated “by compensation a reallocation of the otherwise expected, stress bearing second mora to the preceding drop, resulting thereby in its bimoraic stressed realization” (p. 33). This description of the suspension of resolution echoes the one given in Suzuki (1996) for Old English. As can be seen, the phenomenon is analysed, using linguistic terminology, in particular some sort of moraic theory, and it also reflects the traditional understanding in English historical scholarship of the phenomon of resolu- tion and its suspension. But several questions unavoidably arise here. The idea seems to be that a mora which should normally be required for a regular lift is somehow moved to the preceding drop, which is then heavier and more prominent (cf. Sivers’ idea of Nebenhebung). It is not clear how this relates to the “standard” view within metrical phonology (cf. e. g. Hayes 1995) that moras belong to syllables (or even segments) within stress feet or words. The mechanism of moving these moras between constituents in a metrical line inevitably leads us into the realm of phrasal phonology, where, to my knowl- edge, moras have not been called upon in the same way as within words. And from the point of view of the type-system, one might ask whether a different interpretation would be appropriate. Since (part of ?) the stress or realisation of the lift seems to be moved to the left or into the drop by some sort of syncopation, the obvious question would be whether this compositional type (e. g. Hym 4.7 ástráð mikit, alliterating on the first syllable) should not be moved to the D-class, i. e. / / x x, i. e. with two stresses at the beginning realised by the two constituents of the compound: ást-ráð ‘love-advice’. This of course recalls Bliss’ interpretation (1967: 108) that the “musical” force behind the array of types is based on the displacement of stresses. According to Bliss, the norm for Old English verse is the A-type rhythm / x / x, with an alterna- tion between stresses and non-stresses, but sometimes displacing the stresses to the left (forwards) or to the right (backwards). “If the first stress is placed forwards the result is […] Type C; if the second stress is displaced forwards the result is […] Type E; if both stresses are displaced forwards the result is […] Type B; if the second stress is placed backwards the result is […] Type D”. Thus the variation is based on some sort of syncopation, i. e. variation based on the movement of accents away from their most normal place in the trochaic type. However this may be, suspension of resolution or some sort of displace- ment of stress, compensating for the light lift by a heavy drop, does not tell the whole story about A1s, according to Suzuki. When it comes to lines like HH II 11.6 hildings synir, Grp. 12.5 leið at huga and Hdl. 1.6 ríða við skulum, he 114 Kristján Árnason hesitates to assume heavy drops in the second position (which is what Sievers did, assigning these lines to the A2 subclass A2k). In fact Suzuki argues that in these lines the drop is a normal one (x) and this is why he gives the lines the label A1s; this is then not a variant of A2 with a heavy drop (which was Sievers’ motivation for labelling them A2k), but a subtype of A1. In Suzuki (1996: 82 ff.) several arguments are presented in favour of this position regarding the Beowulf corpus. The main argument seems to be that the distribution and alliteration pattern of these lines are different from A2 proper and more like that of normal A1. Thus on p. 37 we learn that “the lack of significant difference along the two parameters between the two minimal pairs of configurations may justify us in identifying the configuration Px#px as a variant of subtype A1s, and the whole variety of Px…px may by implication be subsumed under the same subtype”. The implication is that in lines like HH II 11.6 hildings synir, Grp. 12.5 leið at huga and Hdl. 1.6 ríða við skulum the positions before the light short lift, is in fact basically a normal drop. Taken literally, this would seem to be a rather radical departure from the Sievers-model. The half-lift or heavy drop, compen- sating for the shortness of the second lift, is no longer necessary, and in general this may seem to open the way for a lift to be interpreted without satisfying the basic requirement of being long or resolved, or suspended from resolution by some special conditions. The argumentation is rather complicated indeed, but the general finding seems to be that this suspension is an optional operation (transformation if you like), which does not create a new type. And the whole thing raises a bigger question, which is the interpretation of the role of length or syllabic quantity in the metre. One may wonder whether the whole structure of the type model as a deductive system is crumbling and falling to the ground. Type A3 A further instance, where it seems that the system is stretched to its limits and away from the original model, is the case of A3. The origin of this type lies in lines with alliteration only in the second lift (Sievers 1885: 11), e. g. a-verses like Vsp. 17.5 fundu á landi, but also b-verses like Vsp. 9.4 ok um þat gættusk. If it is assumed that alliteration is a necessary condition for a lift, such lines would have to do with just one lift, but it seems that this was not Sievers’ understanding, since e. g. he marks Vsp. 9.4 as an A3 type having a resolved (verschleifbar) first position, implying a scansion like |ok um þat |gættusk (Sievers 1885: 18, see also p. 24, where Vsp. 37.1 stóð fyr norðan is 115Prototypes and structures in eddic poetry classified as A3). However, Bliss (1967: 61) characterises A3 in Beowulf as a light type: “Light verses are those which contain only one stressed element” (= Kuhn’s Satzteile). Thus it seems that another important foundation of the typology is no longer sound. And this has led to alternative interpretations of the metre, cf. e. g. Goering’s (2016: 72) conclusion, that “[t]he rhythm of Old English Verse is not “zweihebig”, or indeed “hebig” at all”. In his analysis, Goering adopts Russom’s word-foot model; in other words lifts are not part of the structure. According to Suzuki, the type A3 occurs in all Germanic genres, but in a different shape. In English and Old Saxon, the type is characterised by “demotisation” of the first lift to a drop, whereas in Nordic it is characterised by an “unrealized initial position which must be identified as a lift” (cf. Suzuki 2016: 133). The type is thus characterised in fornyrðislag by the absence of the first lift in its explicit form, since there is only one lift on the surface, located in the penultimate position. The typical configuration is x…Px, in which the last two positions are realized by a simplex word. For Suzuki, this zero realiza- tion is some sort of mirror image or reflection of catalexis, i. e. there is a silent strong, somehow unrealised beat at the beginning of lines, similar to the silent weak one at the end of a catalectic line (giving his types like A1-, C- and D-). Like most deviations from the prototypical trochaic pattern, the A3 type (e. g. Vsp 19.5 þaðan koma dǫggvar; HH 6.1 stendr í brynio; Sg 35.4 riðut at garði) is subject to some distributional restrictions. In West Germanic, the type seems to be banned from the b-verse altogether, whereas in eddic this is only a statistical tendency. According to Suzuki, the relaxation of the cat- egorical prohibition against the occurrence in the b-verse in fornyrðislag can be ascribed to the development in Nordic of a new underlying structure with a lift (albeit unrealised: [/]x / x), which makes the metrical representation conform to the requirement that the b-verse must have two lifts (even if one is empty) and two drops. The difference between fornyrðislag and Beowulf, according to Suzuki, is thus that in the latter, the line begins with a “demoted lift”, by some sort of transformation (SwSw → wwSw), whereas for fornyrðislag there is simply an option for a structure with an empty lift or strong beat. This conclusion is based on the difference between West Germanic and Nordic, that a minimum of two syllables is required before the only lift in the A3-type in the former, whereas just one is needed in Nordic lines like Vsp. 64.4 á Gimlé, Vkv. 9.1 Gecc brúnni [x#Px] (MONEP, p. 60). Such lines are taken as evidence that the first lift may be skipped altogether. But all is not said regarding this type, since the lack of material which would be implied by the absence of a lift is often compensated for (also in Nordic, but obviously not always). Thus the type A3 116 Kristján Árnason requires a “substantially greater number of syllables” than type A1 by way of compensation; what the verse lacks in stress it makes up in length. Needless to say, disbelievers would be likely to raise questions about the limits of the system, or its power of analysis: what sort of transformation defines the relation between basic type / x / x and the variant missing the first lift? As mentioned before, it is suggested that this is in some way similar to catalexis, where a drop is skipped at the end, but there seems to be a substantial difference between a silent strong beat at the beginning and a silent weak one at the end. Of course Suzuki is aware of such questions, and some of them are addressed in the quite lengthy presentation. We cannot go through all of this in this review, nor the details of the motivation for the conclusion at the end of Chapter 2 (pp. l62–163) with the first approximation of an overview of the system of verse types. The bottom line is that we have the traditional five classes A-E, but some of them have subtypes. In particular Class A is listed as having 7 types (A1, A2a, A2b, A3, A1-, A2a-, A3-), Class C as having two (basic and C-), and Class D has three types (basic, D- and D*), whereas Class B and Class E have no variant types. But after having considered anacrusis and catalexis in a special chapter, a revised version is presented on p. 204, basically assuming that the opera- tion of catalexis does not create special types (A1-, A2- and A3-). “Increased” variants are generated by anacrusis and heavy lifts, and reduced ones by “zero realization” of lifts (giving e. g. A3 as a subtype). The final result is, then, that the proliferation of types is controlled or gov- erned by these metrical operations or transformations. Thus the A-type has a basic variant, labelled A1, plus two increased variants: A2a with a heavy first lift, and A2b with a heavy second lift, and furthermore the reduced A3 type. And the catalectic lines are simply classified as subtypes of the respective basic types, so that A1 is the basic one with two subtypes, a catalectic and non- catalectic one etc. What we have is then a conscious attempt at developing a model by which subtypes are defined on the basis of mapping relations based on the basic ones instead of by ad hoc repairs. The metrical mapping: resolution and its suspension This brings us to the next level in the relation between form and text, which is the way in which linguistic form relates to metrical form in the individual types and subtypes. Again, one of the basic questions is whether the same con- ditions prevail in the different cultures, e. g. the English and the Icelandic one. 117Prototypes and structures in eddic poetry Although there is a common thread in the theoretical frame of Suzuki’s work on West Germanic and Norse, it seems that a change can be seen, at least in emphasis, between the treatment of Beowulf (1996) and the present analysis of the Nordic metres. In 1996 the first two chapters are devoted to metre, text and production and the linguistic foundations of metre: linguistic structures (language material in Suzuki’s terms), poetic structures, composi- tion, and performance. In MONED, the emphasis seems, as we have seen, to be more on the statistics and the development of prototypes. The conception of “usage based grammar”, emphasises actual linguistic expressions or “outputs”, rather than underlying structures and derivations when it comes to analysing sentences, words and phonological structure. Thus according to Bybee (2006: 714) “constructions are the basic units of mor- phosyntax”, and grammar is “emergent from experience, ever coming into being rather than static, categorical and fixed”. The idea of metrical prototypes would then seem to imply that the distinction between types and subtypes does not have to be categorical, allowing for alternative interpretations of simi- lar expressions (in this case actual verses and lines in poetry). But the distinction between metrical form and its “language material”, which constitutes the text, is still categorical according to MONEP. And in every instance, principles or rules for the mapping between the “language material” and the metrical form must be followed. Thus throughout the book a clear distinction is made between metrical forms like A1, represented by formulae like / x / x and the linguistic structure or composition of text, rep- resented by formulae like Px#Px (P = “primary stressed long syllable”; x = unstressed short syllable # = word boundary), see e. g. Vsp.1.2 helgar kindir. And other such linguistic strings are defined, e. g. PS#Px (S = secondary- stressed long syllable), e. g. Hym 24.1 Hlǫðvés dóttir, which is metrically analysed as / \ / x (a lift – a heavy drop – a lift – a drop). One of the principles accounting for the mapping between linguistic and metrical form is the above mentioned resolution (German Auflösung), which is said to be a “metrical convention” by which a short linguistically stressed syl- lable and a following unstressed syllable are treated as equal to a long stressed syllable (i. e. ᴗ ́x = ̶́ ). This is exemplified by lines like Grp 16.1 Brotin er brynja (p. 213), which is type A1 with a resolved first lift, and the line has five syl- lables. MONEP (p. 309 ff.) in fact provides an interesting discussion of the conception of resolution, pointing out several unclarities and confusion in the use of the concept in the literature. One such problem regarding the equation two short = one long is its application in connection with what Sievers calls Verschleifung, which Suzuki translates as “slur”. This is, as Suzuki points out, at least sometimes understood 118 Kristján Árnason by Sievers and others as a performance phenomenon involving the contraction or compression of two weak syllables into one in the drop. (This interpretation is then similar to what Snorri Sturluson (1991: 8) calls bragarmál, i. e. when e. g. þar es ‘there where’ is contracted to þar’s in performance.) This is obvi- ously something quite different from the metrical equivalence of two light or short syllables to one long or heavy one in the lift. A typical example of this type of contraction from eddic would be Vsp. 1.3 meiri ok minni, where the final -i of meiri and the o- of ok might be contracted in performance, or the first one deleted, to something like meir’ok minni. Yet, this performance phenomenon has, as Suzuki puts it, sometimes “encroached on the conceptual domain of resolution” through the term verschleifbar referring to two syllables which could function as one position in metre, a lift or a drop. Thus in earlier scholarship resolved drops are sometimes referred to as somehow being on a par with resolved lifts, and we have to agree with Suzuki that this is wrong. In fact there is still another phenomenon, which should be considered in this context, at least when dealing with Icelandic poetry. This is the license of placing two weak syllables in an otherwise regular trochaic pattern without their necessarily being verschleifbar or contractible in performance. This is called neutralization in Árnason (2000 [1991]: 126–130), and an example from dróttkvætt is: |sverðs nema |hefndir |verði. Here one of the weak positions in a basically trochaic line is permitted to have two unstressed syllables instead of one, making an exception in the mora counting structure of the line. There is no reason to assume that in these circumstances a slurring or contraction of the two syllables to make one takes place in performance. (We have no evidence for contracted forms like n’ma or anything similar.) Eddic examples like Vsp. 6.7 undorn ok aptann might fall into this category. This interpretation of the licence of extra, rhythmically neutral, syllables in the drop strength- ens the case for agreeing with Suzuki when he argues against extending the metrical notion of resolution to the drop, concluding that “resolution is […] a privilege of the lift” (p. 213). This also fits the linguistic interpretation that resolution of the strong position is based on disyllabic stress (moraic trochee, cf. Hayes 1995; Allen 1973: 170 ff. talks about disyllabic stress matrices) in words like brotin ‘broken’, whereas the Verschleifung or slur and neutralization allow sequences of more than one syllable to fill the drop without doing much metrical harm. The conclusion is, then, that resolution only takes place in lifts in Icelandic, and the bulk of Chapter 4 (p. 213 ff.) is devoted to a detailed statistical description of the distribution of resolved lifts in the different types in the eddic material. It turns out that there are many statistically significant trends and differ- ences in the distribution of the different configurations. Thus for the type A1, 119Prototypes and structures in eddic poetry resolution is common in the first lift, but quite rare in the second one. For type B, there is only a single example of resolved first lift, none of the second one, or rather it is proposed that candidates for such a scansion should be analysed as A1s (e. g. Vkv. 18.8 æ fjarri borinn, cf. p. 238). The discussion is quite lengthy and, for the present reviewer, often rather difficult to follow. There is thus a rather elaborate discussion about the fact that the first lift of the C-type is commonly resolved but not the second one. It is concluded that first lift of the C-type can be resolved if the second lift is a normal long monosyllable. On the other hand “resolution of the first lift is incompatible with suspension of reso- lution on the second”, since then “relative to the immediately preceding most prominent position in the verse, the following one can hardly be perceived as fully prominent” (pp. 219–220). We learn that “[t]he extra prominence of the first lift through resolution […] enables the perceptually diminished second lift to be realized in relatively stronger shape by virtue of the now increased dif- ferential from the preceding position. As a consequence the cadence Px comes closest to its optimal form in manifestation”. There is thus an inverse relation regarding the realisation of the lifts: “The closer degree of approximation to the prototype of the cadence Px […] favors the configuration x…pxPx over x…PPx, thereby resulting in the conspicuous presence of the resolved first lift in the b-verse […]” (p. 220). And we also see the overarching tendency for the cadence to be regularly trochaic. The general trend shown in Chapter 4.2 (ending p. 241) is thus that the first lift of any type is more prone to resolution than the second one. But there is also a clear difference between the different types in this respect. The two (sub-)types showing the largest proportion of resolved first lifts are type C (36.20%) and the catalectic or shortened D- type (D-, 36.67%). (Compared to this, type A1 has a resolved first lift in only 4.92% of its occurrences.) There is quite a bit of deliberation as to the causes of these statistical differences. For example on p. 230 it is suggested that the first lift is more susceptible to resolution in the C and D- type because of its proximity to another equally prominent lift (since these types, respectively x / / x and / / x, have two adja- cent lifts). And it is said that “[r]esolution effects an increase in prominence through the resulting disyllabicity, which obviously contributes to a sharper differentiation from the following, potentially competing position” (p. 230). The outcome of all of this is that although resolution is initially defined as a simple equation (generative rule if you like) between light disyllables like vini ‘friend-DAT’ and monosyllables like ferð ‘travel’, there are several factors which prevent these forms from being classifiable as totally equal. Several forces seem to interfere and affect the application of the principle. These conditions are quite complex and unclear, and the relation or ‘operation’ of resolution ends up 120 Kristján Árnason looking quite mysterious to the outsider. The idea is of course an inheritance from earlier scholars and has been applied to poetry from different times and cultures, and the questions must be asked, whether it is appropriate to look on it as a coherent concept at all. Do we assume that the “convention” has been passed on as some sort of artistic device, from one generation of poets to another, irrespective of the different linguistic and cultural conditions? As mentioned above, resolution may, under certain conditions, be “sus- pended”, so that lifts are formed without being either long or resolved and thus satisfying the requirement of minimal bimoraicity. In Suzuki (1996: 81 ff.) there is a description of the phenomenon, which is said to be condi- tioned by the strength or weight of neighbouring syllables. According to this interpretation “the non-occurrence of resolution is compensated for by the stress on the preceding drop” (Suzuki 1996: 85). And “the demand for a long stressed syllable is carried over to the first drop, which in turn is required to be matched with the remaining stressed mora”. In some way the second lift is exempted from filling the requirement of having two moras, “and the imme- diately preceding metrical position (the first drop) is instead invoked for its satisfaction. A stressed mora, however, is incapable of occurring in isolation. This amounts to the stipulation that the first drop be filled by a long stressed syllable, the second mora of which serves as a substitute for the missing mora of the penultimate syllable associated with the second lift. Thus “the sequence of a long stressed syllable and a short stressed syllable is made equivalent to the sequence of an unstressed syllable (of arbitrary length) and a long stressed syllable” (ibid.). This formulation of the compensation of the shortness of the lift by the extra mora in the drop is echoed in MONEP (p. 33) in a statement to the effect that the realization of a lift on a short syllable “motivates by compen- sation a reallocation of the otherwise expected stress bearing second mora to the preceding drop” (p. 33). Examples of such lines are: Hym 4.7 ástráð mikit. So, the drop is somehow made responsible for realising the lift, which is (metrically?) assigned to the short syllable. Although not stated clearly, this motivation would seem to have to be some sort of performance phenomenon or a phrasal sadhi, connecting moras from different words. And some sort of movement or shift of prominence from right to left would seem to be involved. An obvious question here seems to be whether we should not assume an allu- sion in the poetic text to a SSvv rhythm (i. e. with inversion of the type found in dróttkvætt: (cf. gnýskerðandi verða; hjálmfylli spekr hilmir, Árnason 2000 [1991]: 124 ff.). In fact the moraic interpretation of the suspension of resolution seems to be played down in the discussion in MONEP, and in some way the road 121Prototypes and structures in eddic poetry is opened for short syllables more generally forming lifts on their own, i. e. without the support of moras. Thus on p. 258 we learn that lines ending in words containing short stressed syllables “are all suspended from resolution in such a pervasive and consistent way in fornyrðislag that implementation of resolution […] is categorically excluded as an alternative metrical operation”. This leads to the following rule stated on p. 259: “The short stressed syllable must constitute a lift on its own only when it is followed by the verse final strings –x# or -Xx#”. Irrespective of how this is to be interpreted exactly (is this an “obligation” or a “permission”), it implies that the conditions for the application or suspension of resolution in Icelandic are quite different from (and even more complicated than) the conditions in Old English. One of the complications in the workings of resolution is Kaluza’s Law. In Old English, this law defines a restriction to the effect that under certain conditions a disyllable is resolved only when it ends in a short vocalic ending derived from a pre-Old-English short vowel (PGmc. *i or *u). A resolved form has to have some stress (primary, secondary or tertiary?). This phenomenon gets its share of discussion in MONEP and it is said to constrain the operation of resolution, so that “after a stressed syllable, only the least sonorous disyl- lable [i. e. consisting of a short stressed syllable followed by a short unstressed one] was qualified for resolution and hence constituted a single position” (pp. 305–306). This is basically equivalent to saying that, at least under certain conditions, the resolved form must be strictly bimoraic, i. e. having two short syllables. The long discussion which follows concludes that although there are some indirect correspondences, the metrical and linguistic conditions are dif- ferent, and the law cannot be applied in the same way in Nordic as in English. Metrical innovations: málaháttr and ljóðaháttr The final chapters of Part I dealing with fornyrðislag contain additional obser- vations and statistics regarding the emergent tendency in Nordic for regularity in the cadence, taking the form of lift + drop (with catalexis as an option), the distribution of alliteration, and the line-types within the stanza. Recognizing the stanza as a “higher metrical unit over the line”, Chapter 7 is devoted to a thorough statistical overview of the distribution of textual differences or types within the stanza (verse 1–8), and similarities, e. g. between verse no. 1 and no. 5. This is all quite informative, but it seems a bit surprising that more attention is not paid to the fundamental question of how and why stanzaic structure developed in Nordic. The 8 verses and four long lines obviously form a 4 × 122 Kristján Árnason 4 structure, which as I have pointed out, makes the eddic stanzas look like quatrains counting phrasal stresses. And this fits well with the development of ljóðaháttr as a special type of quatrain with truncation (skipping a foot) in even numbered lines (cf. Árnason 2006). Dealing with related issues, the last two main parts of the book are devoted to the specifically Nordic metres, málaháttr (Part II, Chapters 8–11) and ljóðaháttr (Part III, Chapters 12–14). The treatment is similar to that of fornyrðislag: after rather superficial remarks regarding the basic differences of these new metres compared to the fornyrðislag, statistical overviews are given of the compositional types. The málaháttr is simply said to have five positions per verse, compared to the four of fornyrðislag, being expanded with an extra drop. A further feature of this metre is that the trochaic cadence / x is “maxi- mized in its organizing power so that the two verse classes – B and E – … are almost entirely eliminated from the inventory of legitimate classes” (p. 429). As is well known, the status of málaháttr as a separate form basically depends on just one poem, Atlamál hin grænlenzku, whereas three “peripheral variants” of the metre, illustrated respectively by Atlakviða, Hamðismál and Hárbarðzljóð, are also described in MONEP. Here the question obviously rises, how stringent the metrical definitions are. An alternative might be to include these forms under the same heading as fornyrðislag, for example if it could be shown that they all have two lifts to a line, but obviously this would lead to a still greater proliferation of types for the model to handle. The treatment of ljóðaháttr follows a similar pattern: a brief general char- acterization is given of the main features, i. e. two verses (a- and b-) forming a long line, and a “full line”, labelled c-verse; the main text is devoted to a statistical description of the composition. This is done in two chapters, one devoted to the pair of verses (a- and b-) forming the long line (Chapter 12), and another (Chapter 13) to the c-verse. The treatment of ljóðaháttr ends by a clear and useful overview (Chapter 14) of the stanzaic structure of the metre. This involves both an account of the varying regularity of individual poems and of the different compositional types. And the prototype model once more forms the backdrop of the analysis, so that the six line stanza, divided into two half stanzas is seen as prototypical, but various divergent actualisations of the form are allowed for. All in all this turns out to give a clearly presented and useful overview of the whole corpus. The final chapter, no. 15, entitled Conclusion, at the same time as summarising the general findings, presents an “evolutionary trajectory of the eddic metres as they divergently developed from their Common Germanic and North Germanic Ancestors” (p. 773), cre- ating the new forms málaháttr and ljóðaháttr beside the basic fornyrðislag. 123Prototypes and structures in eddic poetry Like other parts of the book, this last chapter is “Wagnerian” in length and form. It starts its 26 pages by considering the effects of the specifically Nordic linguistic development, and then moves on to stage the emergence of the new forms. The reduction in the number of syllables in words has an effect on the correspondence between linguistic syllables and metrical positions, moving closer to one-to-one correspondence, and at the same time there is a “decreased distinguishability of the two opposite metrical positions, the lift and the drop”. One effect of these changes is the “removal of the heavy drop” from the D-type. This lead to a “weakened identity” of the heavy drop, and at the same time resolution “becomes closest to being moribund”, although it operates “more vigorously” in certain contexts, e. g. in the C-type (p. 773–774). This has the effect that the constraint for lifts to be heavy (i. e. be filled by long syllables) is relaxed to a great extent, (since e. g. the novel configuration Px…px, with a short second lift becomes a variant of the basic A1-type). Thus there is a proliferation of short lifts. Another development is that the “four-position principle was strictly obeyed in the Norse meter [i. e. fornyrðislag]”. One of the effects of this is that “anacrusis is allowed to occur only on the shortest and least prominent variants of type A1” (p. 777). Adherence to this principle is also seen as responsible for the restriction on expanded types like D* and the development of catalexis. The Nordic syncope leads to “a massive emergence of verses […] which appar- ently end in lifted words” (p. 778). This leads to the reinterpretation of the change in linguistic structure as a metrical principle of catalexis, namely the option for A1- as a variant of trochaic A1 etc. This means that a metrical posi- tion “a verse final drop is actually there in underlying representation and that it is simply realized as zero on the surface through alignment to null language material” (p. 779). Another instance of this sort of null realization of linguistic material gives type A3, “which leaves the initial lift immaterialized”. These new phenomena interplay with traditional “conventions inherited from the Old Germanic metrical tradition”, according to which positions at the beginning of a verse count as more prominent, making the “b-verse […] unmarked in opposition to the a-verse” and the principle that “increase in prominence of a verse promotes use of double alliteration in the a-verse” (p. 780). One effect of this is that catalectic lines are equipped with a lower prominence and therefore tend to occur at the end of lines. One more event in this history is that the “increasing standardization of the cadence heightened awareness of the line as a higher unit over the verse” and this also contributed to the “the emergence of the stanza” (p. 784) as a significant unit. But, like in a Wagnerian Handlung, there are some disturbing and even contradicting forces affecting the development, which among other things lead 124 Kristján Árnason to the creation of two new and quite different metres. Although málaháttr and ljóðaháttr go in different directions in their development and structure, their development is at least partly due to the same impetus, which comes from West Germanic. This would then be something of the sort that was seen by Hans Kuhn (1939) as marking the Fremstofflieder, involving greater variation in the length of lines, but most importantly the non-application of his own, Kuhn’s Laws in some fornyrðislag poems and those composed under ljóðaháttr. For the málaháttr the innovation was the addition of the fifth position as a legitimate one in the line. For the ljóðaháttr it was the other way around, since the innovation involved shortening, i. e. increase in the incidence of shorter lines (a- and b-verses), but at the same time the development of the longer c-verse, which actually is a shortened long line. The two metres also went in different directions regarding the cadence, málaháttr favouring the trochaic ending, but ljóðaháttr the catalectic one. The result of all of this is the above mentioned trajectory for Nordic metrical development (fornyrðslag > málaháttr > ljóðaháttr, p. 797). This projection is meant to be interpreted temporally, so that first the málaháttr developed (through West Germanic influence), and then the ljóðaháttr, making different use of the same influence. We must agree with the implication that ljóðaháttr is not an archaic metre as had been suggested by some scholars. But the way in which West Germanic influence was responsible for the metrical innovations is not too clear. An alternative account of the development of ljóðaháttr, as well as the general introduction of the stanza, is to see it as due to the 4 × 4 principle. This implies that the eddic stanza is a quatrain, made of four four-beat lines (cf. Árnason 2006). This innovation might well, through its naturalness, be home spun, or due to influence from younger European genres. Conclusion The body of work on Old Germanic poetic forms is immense, and for the outsider it may look like an impenetrable jungle where all sorts of rules and principles have kept scholars busy devising complicated rules, types and sub- types. Although there are some basic theoretical issues which can be focussed on, many of the fundamental questions connected with the material seem often to have been forgotten or shrouded by attention to all kinds of detail. And it remains an open question to what extent the whole corpus of Germanic verse, English, German and Icelandic, can be subsumed under one heading. 125Prototypes and structures in eddic poetry In these circumstances, the resilience of the Sievers-typology derives from its usefulness as a pre-theoretical taxonomic tool, which is precisely the origi- nal characterization given to it by its founder, although later scholars have tended to reify the types as some sort of axiomatic formulae (reminiscent of generative syntax). Suzuki’s work fits into this tradition with interesting additional twists by invoking insights from generative phonology and the con- ception of prototypes. But this is not simple, and there would seem to be some problems or even contradictions because of the differences between the two theoretical frameworks. The idea of prototypes is more salient in MONEP than in older work by Suzuki. But the conversion is only half-hearted, because there is still talk of underlying structure: the metrical system and the linguistic one are kept apart, so much so that empty categories (lifts or drops) are allowed without being filled by linguistic material. One may wonder how that would fit the general idea or theory of prototypes, usage based grammar or exemplar theory, as defined by the proponents of that theory. Do empty positions or abstract, underlying structures fit at all into that framework? The value of MONEP lies primarily in that it is a well-organized and thor- ough presentation of the compositional characteristics of Old Norse eddic poetry compared to its West Germanic relatives. We are constantly reminded of the theoretical grounding of the account, and the statistical analysis is very informative, if rather lengthy at times. This is not (to the present reviewer at least) an easy book to read; it takes quite an effort to discern the general patterns, and the wood is often quite dif- ficult to see because of the close attention paid to the endless number of trees of types and subtypes. But in spite of the overwhelming detail and lengthy deliberations, MONEP has a clear focus, and it raises many issues which will help us on the way to fuller understanding of the metrical and linguistic foun- dations of Old Norse poetic forms. The careful presentation of the material, the book’s very professional production, layout and organization make it a most reliable companion to the study of Nordic metrics. But for all its mer- its, like many other works in the field, it seems to fall short of reaching the more explanatory goals foreseen by Eduard Sievers regarding the musical or rhythmic foundations of the forms, their relation to performance, as well as linguistic structure. This is perhaps just as well, since otherwise a number of very learned scholars might be out of a job! 126 Kristján Árnason References Allen, W. Sidney 1973. Accent and Rhythm. Cambridge: Cambridge University Press. Árnason, Kristján 2000 [1991]. The Rythms of Dróttkvætt and Other Old Icelandic Metres. Reykjavík: Institute of Linguistics. Árnason, Kristján 2006. The rise of the quatrain in Germanic: Musicality and word based rhythm in Eddic metres. In: Dresher, Elan; Friedberg, Nila (eds.), Formal Approaches to Poetry: Recent Developments in Metrics. Berlin: Mouton de Gruyter, 151–169. Árnason, Kristján 2007. On the Principles of Nordic Rhyme and Alliteration. In: Arkiv för nordisk filologi 122, 79–114. Árnason, Kristján 2016. Text and form in the Eddic metres: Four lifts and how many types. In: Árnason [et al.] (eds.) 2016, 63–115. Árnason, Kristján; Carey, Stephen; Kim Dewey, Tonya; Þorgeirsson, Haukur; Aðalsteinsson; Ragnar Ingi; Eyþórssons, Þórhallur (eds.) 2016. Approaches to Nordic and Germanic poetry. Reykjavík: University of Iceland Press. Bliss, Alan Joseph 1967. The Metre of Beowulf. Oxford: Blackwell. Bybee, Joan 2006. From usage to grammar: The mind’s response to repetition. In: Language 82(4), 711–733. Cable, Thomas 1974. The Meter and Melody of Beowulf. Urbana, Chicago, London: University of Illinois Press. Fulk, Robert Dennis 1992. A History of English Meter. Philadelphia: University of Pennsylvania Press. Fulk, Robert Dennis 2016. Eddic metres. In: Larrington, Carolyne; Quinn, Judy; Schorn, Brittany (eds.), A Handbook to Eddic Poetry: Myths and Legends of Early Scandinavia. Cambridge: Cambridge University Press, 252–270. Goering, Nelson 2016. The Linguistic Elements of Old Germanic Metre: Phonology, Metrical Theory, and the Development of Alliterative Verse. D. Phil. Thesis. The University of Oxford. Hayes, Bruce 1995. Metrical Stress Theory: Principles and Case Studies. Chicago: The University of Chicago Press. Hayes, Bruce; McEachern, Margaret 1998. Quatrain form in English folk verse. In: Language 74, 473–507. 127Prototypes and structures in eddic poetry Heusler, Andreas 1969 [1889]. Der Ljóðaháttr: Eine metrische Untersuchung. In his Kleine Schriften II. Berlin: De Gruyter, 690–750. [First published in: Acta Germanica 1(2), 89–174.] Kuhn, Hans 1939. Westgermanisches in der altnordischen Verskunst. In: Beiträge zur Geschichte der deutschen Sprache und Literatur 63, 178–236. [Reprinted in Kuhn 1969, 485–527.] Kuhn, Hans 1969. Kleine Schriften I. Berlin: De Gruyter. Larrington, Carolyne; Quinn, Judy; Schorn, Brittany (eds.) 2016. A Handbook to Eddic Poetry: Myths and Legends of Early Scandinavia. Cambridge: Cambridge University Press. Russom, Geoffrey 1998. Beowulf and Old Germanic Metre (Cambridge Studies in Anglo-Saxon England 23). Cambridge: Cambridge University Press. Russom, Geoffrey 2009. Why there are Three Eddic Meters. In: Kilipö, Matti; Kahlas- Tarkka, Leena; Roberts, Jane; Timofeeva, Olga (eds.), Anglo-Saxons and the North. Tempe, AZ: Arizona Centre for Medieval Studies, 69–88. Sievers, Eduard 1885. Proben einer metrischen Herstellung der Eddalieder. Tübingen: Königliche Eberhard-Karls-Universität. Sievers, Eduard 1893. Atlgermanische metrik. Halle: Max Niemeyer. Snorri Sturluson 1991. Edda, Háttatal. Edited by Anthony Faulkes. London: Viking Society for Northern Research. Suzuki, Seiichi 1995. Resolution and mora counting in Old English. In: American Journal of Germanic Linguistics and Literatures 7, 1–28. Suzuki, Seiichi 1996. The Metrical Organization of Beowulf: Prototype and Isomorphism. Berlin: Walter de Gruyter. Suzuki, Seiichi 2004. The Metre of Old Saxon Poetry: The Remaking of Alliterative Tradition. Cambridge: Brewer. Suzuki, Seiichi 2008. On the Emergent Trochaic Cadence / x in Old Norse Fornyrðislag Meter: Statistical and Comparative Perspectives. In: Journal of Germanic Linguistics 20(1), 53–79. Suzuki, Seiichi 2009. Three-position verses in Old Norse fornyrðislag meter: Statistical and comparative perspectives. In: NOWELE 56, 3–40. Suzuki, Seiichi 2010. Anacrusis in Eddic meters Fornyrðislag and málaháttr: Reevaluation and Reinvigoration. In: Beiträge zur Geschichte der deutschen Sprache und Literatur 132, 159–176. 128 Kristján Árnason Suzuki, Seiichi 2011. Catalexis, Suspension of Resolution, and the Organization of the Cadence in Eddic Meters. In: Lotman, Mihhail; Lotman, Maria-Kristiina (eds.), Frontiers in Comparative Prosody. Bern: Peter Lang, 373–400. Suzuki, Seiichi 2014. The Meters of Old Norse Poetry (Ergänzungsbände zum Reallexikon der Germanischen Altertumskunde 86). Berlin: Walter de Gruyter. Suzuki, Seiichi 2016. Toward a formal account of a3 in fornyrðislag. In: Árnason [et al.] (eds.) 2016, 117–145. Terasawa, Jun 2011. Old English Metre: An Introduction. Toronto: University of Toronto Press. Yakovlev, Nikolay 2008. The development of alliterative metre from Old to Middle English. D. Phil. Thesis. The University of Oxford.