1. Introduction.
1.1. Overview.
Runological research usually stresses the exclusiveness of runic writing, in particular when it comes to an assessment of the runic inscriptions of the older fuþark up to 700 C.E. The general expectation of “high” style or, as Beck (2001) puts it, “Stilisierung als ein Wesensmerkmal der Runenkunst,” [stylization as a characteristic of runic craft] derives from assumptions about elaborate metrical forms. The Gallehus inscription, for instance, is an alliterating poem of the structure h-h-h-t = a-a-a-x, the most common type of alliteration in Old Germanic (Watkins 1995:23, Naumann 1998:702). This focus on particular types of texts with stylized forms and elaborate metrical structures results in a general neglect of features of vernacular speech. According to Braunmüller (forthcoming), for instance, “there is no evidence that runic inscriptions in the older fuþark represent actual speech in one way or other.” However, general statements of this kind tend to overshoot the mark, and Derolez (1998a) is concerned with the representativeness of the corpus of older runic inscriptions. Supporting the standard view is the fact that the extant runes tend to be carved in durable material, which is assumed to favor a particular “high” style. By contrast, runes carved in the early period on more perishable (basically organic) material such as wood, birch bark, antlers, bones, or sandstone have been almost entirely lost. If there were early runic inscriptions on organic material, this would allow for the possibility of greater spontaneity and informality in runic writing. However, since such inscriptions would be lost under normal circumstances, as a matter of chance we would have fewer examples of “lower” stylistic properties.1
For a stylistic description in terms of allegro versus lento, see section 1.2. More trivial everyday messages are found with the later runes. On the medieval inscriptions from Bryggen in Bergen, see Liestøl 1964, 1974, 1976, Düwel 2001:156–160, and Spurkland 2005; on the Ladoga inscriptions from the late ninth century, see Høst 1960, Krause 1960, and Liestøl 1970; on the related issue of runic graffiti, see Hammarberg and Rispling 1985, though this article is highly speculative, and Düwel 2001:126 with further literature.
In addition, it is possible that runemasters carving in stone or incising in metal might have worked from a scratch, for instance, a wax tablet. I owe this con-sideration to one of the anonymous reviewers.
Research carried out by Biber (1988, 1995) emphasizes that no absolute spoken versus written distinction is identifiable, and that the interrelations between these two media are complex (see also Chafe 1982). Rather than following a strict dichotomy, we are facing oral–written continua with certain overlaps. In comprehending the development of runic writing from the first century onward as a bifurcation process with the emergence of different text types, the need for stylistic variation (including “lower” styles) seems obvious. The two basic claims of this paper, therefore, are (i) that runic writing receives no special status as compared to other writing systems, and (ii) that runic texts do not constitute an exception to the general rule of textual variation in the sense that they represent only one stylistic register. On closer inspection, there are several indications of actual speech phenomena (including features of allegro styles as defined in section 1.2) even in the older runic inscriptions, particularly in the transitional group of Blekinge.3
There are two basic criteria for assigning transitional status to these inscriptions of the sixth and seventh centuries. First, from a linguistic per-spective, they capture Nordic sound changes such as phonemic umlauts, breaking, and syncope. Second, from a graphological perspective, they reflect the transition from the older 24-grapheme fuþark to the later 16-grapheme fuþark, with the reanalysis and/or exclusion of several runes. See, for example, Barnes 1998 and Schulte, forthcoming a, b.
A central case is provided by the Blekinge curse formula, which is extant in two different versions, namely Stentoften and Björketorp from around 600–650 C.E. As I demonstrate below, these texts show diverse features of oral speech, both linguistically and textually. The reason is that the spoken incantation is vital to its taking effect, hence the close relation of the curse formula to actual speech (see section 5). Frankfurter 1994:195 contains the following remark in connection with Greek amulets:
The formulation of the spell directly reflects oral utterance—“I bind NN;” “Restrain NN!”—and that the preparation of the medium itself (a lead tablet) derives from gestures to render the victim “like” lead. Thus again, the ritual performance […] dominates the written spell; the written spell essentially “records” the ritual. These inscribed spells carry the implication that the initial ritual “speaks on” through the written word […].
Although runic curses cannot directly be compared to Egyptian and Greek incantations, the focus on the spoken word is clear. The written text is intended as a mapping of the oral utterance, which itself guarantees the efficiency of the formula, and the accurate recitation is thus of vital importance (Düwel 1997:28, 32). Due to their affinity with oral speech, runic curse formulas (often labeled “magic runic inscriptions”) are expected to show characteristics of spoken language. This observation forms the basis of the present analysis.
1.2. Outline and Theoretical Framework.
Stylistic inquiries play a central role in the present study of older and younger runic inscriptions. This is true both for the textual and phonological data. Examining the textual data, the focus is on elliptical shortenings and condensations, whereas the investigation of the phonological data focuses on non-standard spellings that are conditioned—at least partially—by actual speech (see Spillner 1984:79–90). Such spelling variations can be due to a number of reasons, such as conscious or unconscious stylistic adjustment (see section 5).
Following Dressler and others, the present study invokes a theoretical framework of fast-speech phonology, in which LENTO versus ALLEGRO styles—forming the ends of a continuous scale—can be given the definitions below:4
See Dressler 1972, 1973, 1974, and 1975a, Dressler and Wodak 1982, Shockey 2003:113–114, and Zwicky 1972:§1.
LENTO: A rather slow and conscious (even unnaturally accurate) articulation of single words as full forms that are uttered separately or in word group and structured by clear pauses without far-reaching contractions or assimilations involving external sandhi rules.
ALLEGRO: A rapid, rather careless articulation of sequences in connected speech involving shortenings and far-reaching sandhi phenomena (in particular the loss of juncture, assimilations, and contractions). The crucial factor is carelessness rather than speed, hence the focus on casual speech phenomena.
This stylistic definition in terms of allegro versus lento provides a framework for identifying casual speech phenomena in the transitional inscriptions of the seventh century and in the later Viking Age inscriptions. Hallmarks of vernacular speech to be evaluated are the epenthetic vowels (section 3), the neutral schwa values in unstressed syllables (section 4), and the above-mentioned textual condensations signaling the influence of less formal, allegro, styles (section 5). Also investigated in this framework of allegro phonology is whether processes of connected speech (fast-speech rules) are reflected at the sentence level in runic epigraphy (section 6). Finally, I investigate whether the notion of allegro forms sheds new light on the problem of early sound change (section 7). This is particularly relevant with respect to those inscriptions where the above-mentioned speech phenomena have already been traced.
2. Sporadic Oral Traces in Early Runic?
As a starting point, I turn to the question of whether there are early traces of oral influence on runic writing. A rare example (circa 400 C.E.) involves the neutralization of the voice contrast /t/:/d/ after /s/ in the second element of asu-gasdiz on the Myklebobostad stone (KJ 77) as opposed to hlewa-gastiz on the Gallehus golden horn (KJ 43; 400–450 C.E.), reflecting PG +gastiz, ON gestr ‘guest’.5
See Krause 1971:§20, Antonsen 1975:§4.6, and 2002:9–10. Incidentally, the Björketorp form sb A (KJ 97; 7th century) in relation to ON sp
‘prophecy’ (PG +spahu), though mentioned by Williams (1992:203) in this connection, provides no clear evidence due to the early replacement of the rune p by b in the older runic inscriptions. On the marginal status of the p -rune in the older fuþark, see Odenstedt 1990:79, 93–94, among others. Note that KJ followed by the number of the inscription refers to Krause and Jankuhn 1966.
For a parallel instance in South Germanic epigraphy, see the belt buckle of Weimar II where the sequence isd is interpreted as 3rd person singular present indicative ‘is’ (auxiliary verb), phonological form /ist/ (Nedoma 2004:350–351):
Dabei ist wohl anzunehmen, dass in gegenständlichem isd die phonologische Opposition zwischen stimmhaftem und stimmlosem dentalen Plosiv in der Position nach /s/ neutralisiert ist und das “Archiphonem” durch “d” bezeichnet wird. 6
“It must be assumed that the form isd conveys the neutralization of the phonological opposition between voiced and voiceless dental plosive and that the ‘archi-phoneme’ is designated by ‘d’ in this case.”
Hence, the spelling d instead of t can be phonetically justified. Indeed, the same factors are evident, for example, in children's English spellings such as sbun ‘spoon’, sdov ‘stove’, sgie ‘sky’, etc. (Treiman 1993:144–148). Derolez (1998b:113) mentions Old English word pairs that attest to this phonotactic rule, such as bere ‘barley’: spere ‘spear’, brecan ‘break’: sprecan ‘speak’; see also ON barða: sparða ‘(two different types of) battle-axe’. As the direct evidence for oral features of this type in the oldest runic inscriptions is extremely scarce, I turn to a phenomenon of considerably wider range.
3. Vowel Epenthesis and Speech Tempo.
The argumentation developed in this section is based primarily on the evidence deriving from epenthetic vowels in runic inscriptions. The type of insertion where a vowel breaks up a consonant cluster, in particular a liquid cluster, is known as anaptyxis or vowel epenthesis; for example, +wrīt- ‘write’, which is represented in KJ 70 Järsberg stone (Sweden; 500–550 C.E.) as waritu, in KJ 98 Istaby stone (Blekinge, Sweden; circa 650 C.E.) as warAit, as opposed to KJ 17a Eikeland fibula (Norway; circa 600 C.E.), where it appears as writu, and probably KJ 74 Reistad stone (Norway; 350–500 C.E.), where it appears as wraita (see Schulte, forthcoming c). This extremely common crosslinguistic process can be regarded as an operational rule in the older runic inscriptions.
I begin with three basic observations. First, as Makaev (1996:51) remarks, there is a clear increase in this phenomenon in Transitional Runic (especially in the Stentoften-Björketorp group) and again later in the inscriptions of the Viking period (DR 1003–1005). Second, epenthetic vowels are not only found in runic inscriptions, but are characteristic of other Germanic and non-Germanic languages, including later stages of the Nordic languages, for example, Modern Icelandic hestur < ON hestr ‘horse’. Third, terms such as parasite vowels or Lemuren (Krause 1971:83) are probably unsuitable, since these additional vowels are structurally-based synchronic markers, indicative of phonological changes in the speech syllable. In this regard, Makaev (1996:51) raises the question as to “whether epenthetic vowels should be considered a determining feature of runic graphics, or simply one of the characteristic features of graphic typology in many ancient languages, which only partially influenced the principles of runic graphics.”
Under the present analysis, vowel epenthesis is considered a determining feature of Transitional Runic. Traditional accounts relate epenthesis in runic epigraphy to a lento articulation of connected speech, which thus would be due to the slowness. This general argument links the rune carver's articulation to the notoriously long-lasting process of incising runes in stone and other materials. In the words of Liberman (2000:113):
Runemasters carved runes and, therefore, could not “write as they spoke.” The numerous epenthetic vowels in runic inscriptions have hardly been accounted for better than by reference to the laborious process of carving.
Though this standard account seems convincing at first sight, it may be challenged on several grounds. In my view, it is widely speculative to draw direct inferences from the long-term writing process in hard material to the phonological style of the represented text. Being highly context-sensitive, linguistic style depends on a variety of factors such as text type, degree of formality, message, and the social status of the language user and the addressee.7
See section 1.2 above. For a general assessment of style in a framework of socio-historical linguistics, see Romaine and Traugott 1985.
It may further be argued that epenthetic vowels in Transitional Runic are a hallmark of rapid speech.8
See also Schulte 2005a:15–18 and 2005b:173–174.
This view [the traditional one based on lento speech, M.S.] is unsupportable for a number of reasons. In the first place, a rapid or normal speech tempo seems a priori more likely to result in epenthesis. Transitions in phonotactically difficult sequences will only become more problematic as speech tempo increases. An epenthetic vowel inserted in an original sequence of CVrC > CVr
C as in berht > bereht brings the syllable structure closer to the universally preferred CV.CV, a syllable structure undoubtedly better suited to rapid speech. Secondly, it is difficult to imagine that all of the epenthetic vowels that occur in Runic inscriptions, Old High German, Old English, Middle Dutch, and elsewhere result from scribes slowly pronouncing words to themselves in hopes of arriving at an adequate spelling. The different range of environments for epenthesis in, for example, Franconian and Upper German dialects of Old High German implies that these epenthetic vowels represent a regular phonetic characteristic of these dialects. Finally, modern dialect studies show numerous examples of vowel epenthesis in precisely the environments where such developments are reflected in the early Germanic sources. What is more, they occur with great frequency in normal or rapid speech in many speakers of modern standard Dutch.
Following Howell, epenthetic vowels simplify the output structure and thus render the articulatory process more suitable for rapid speech. In support of this, Hock (1991:125) notes that “in the majority of cases, the motivation of anaptyxis seems to lie in the elimination of consonant clusters as an alternative to cluster simplification.”9
The historical development of the cluster -lh- in PG +felhan- > ON felasupports the close functional parallel between vowel epenthesis and cluster simplification (Howell 1991:75–76). Conversely, vowel deletion due to fast speech yields new consonant clusters, for instance, French p(e)tit, English p(o)lice.
So far, I have confined the discussion of spoken features to epenthesis at the word level. To support the claim that runic inscriptions reflect actual speech phenomena, let us now turn to the general representation of unstressed vowels in Transitional Runic.
4. Unstressed Schwa Values.
A directly related and generally ignored problem concerns the representation of schwa values in transitional inscriptions around 600–650 C.E. Again, this characteristic feature indicates a close interrelation between oral and written language. In runological analyses, the vowel entity schwa is often ignored due to its unclear phonemic status. For example, Antonsen (1975:16–17) and Nielsen (2000:96–97) do not mention it in their phonological accounts of the unstressed vowel system of Transitional Runic. However, Boutkan (1995:37) and Birkmann (1995:134) both stress the need to posit schwa in the Blekinge inscriptions.10
From a general perspective, orthographic systems typically do not follow phonemic principles consistently; see Miller 1994:xii–xiv: “Writing systems are attempts at representing different, ‘competing’ aspects of language […], some phonetic (non-contrastive), some phonemic (contrast and opposition), some lexical/morphological (root or affix unity), some morphophonemic (in the broad sense). Such competing goals are apt to yield discrepancies and irregularities in graphic conventions” (p. xiv).
Every linguistic assessment of a given (also runic) text demands “an exact linguistic analysis with respect to the sound system of a given language” (Jakobsen and Waugh 1979:216). Focusing on the Nordic syncopation laws, Boutkan observes that some vowels designated by A, a in the Blekinge group, are subject to syncope in later Old Norse, while others are not, for instance, runAR (feminine accusative plural) < PG +ru
n
z, ON rúnar as opposed to hAriwulafa (accusative singular) < PG +-an, ON Heriólf (both forms occur in the Istaby inscription KJ 98 circa 650 C.E.). Boutkan (1995:37) states:
If the earlier long vowel +ō had developed into an E[arly] N[orse] short vowel /-a/ that remained in O[ld] Ic[elandic], the phonetic quality of the earlier short vowel -a < -an must have differed from this new /-a/ during the period that both sounds coexisted. It may have been a reduced vowel
.
This diachronic argument is obviously a strong one, and my own analysis of the transitional inscriptions corroborates this view (Schulte 2003:393–394). Such schwa values came into being through gradual weakening processes and must have disappeared in the course of syncope (that is, via schwa deletion). Graphologically, reduced endings are denoted by the graphemes A, a, which convey the idea of centralizations and mergers of the originally distinct theme vowels +i, u,a about 600–650 C.E. (Schulte 1998:152–158). Runic examples from the Blekinge group and the Ellestad inscription are given in 2.
For detailed discussion, see Schulte 1998:76–158, 2003:393–394, and Rasmussen 2000:147–148. On the phonological form of Ellestad KJ 59 sigimArAR, see Schulte 2004:85–88.
This phonetic assessment supports the basic claim of this paper that runic inscriptions show features of vernacular speech. By means of sound-analysis, the runographers of Stentoften, Björketorp, and Ellestad (circa 600–650 C.E.) establish basic correspondences between the graph-emes A, a, and schwa vowels, such that they turn away from stable word-images with traditional spellings and rely on their own phonetic analysis of vernacular speech instead.
5. Elisions and Ellipses.
Let us now investigate the texture of the Blekinge curse formula of Stentoften and Björketorp with respect to omissions and ellipses. Elliptical constructions, an essential feature of everyday speech, are sometimes referred to as reduced, contracted, or abbreviated structures. For reasons of style, emphasis, or economy, a part of the underlying structure may be omitted, which is transparent in the grammatical struc-ture or recoverable from a scrutiny of the context. Crosslinguistically, unstressed function words such as the copula or relative particle are particularly prone to reduction and elision (for copula deletion in English dialects, see Labov 1969, Mesthrie, Swann, Deumert, and Leap 2000:365–366.
As for Transitional Runic, the determination of styles yields new insights into the texture of central inscriptions such as the Blekinge group. The issue of style aims at the textual structure (see Romaine and Traugott 1985), or as Beardsley (1981:222) puts it, “the recurrent features of the texture of meaning.” A case in point is the curse formula of the Blekinge group, which is extant in two different versions, namely, Stentoften in 3a and Björketorp in 3b. Apart from systematic spelling divergences, the two parallel texts exhibit different grammatical and morphosyntactic structures, or as Nedoma (2005:171–172) would have it, “zwei verschiedene ‘Formulierungsverfahren’ […] mit Verschieden-heiten in der Textkonstitution” [two different wordings […] with differences in textual constitution] as displayed in 3. The rune sequences missing in 3a as compared to 3b are underscored.
As Grønvik (1996:157) notes, the general judgment about the relation between 3a and 3b is largely based on Krause 1966:217, who regarded the Stentoften text as a corrupted copy of Björketorp with several inadvertent omissions of rune sequences (that is, the underlined sequences in 3b in relation to 3a). This general reliance on Björketorp is also expressed by Antonsen (2002:307), who states that “Stentoften's … no must be a writing error in which the first two runes were inadvertently omitted.”
Other scholars such as Snædal (1997) and Barðdal (1998) subscribed to this view, leaving the Stentoften text entirely out of sight. But their isolated reading of the Björketorp sequence utiAR as the 3rd person singular present of the weak ON jan-verb vitja ‘haunt’ is already extremely problematic on epigraphic grounds.12
See Snædal 1997:155 and Barðdal 1998:12–15, and for criticism, Nedoma 2005:182, note 59.
On the linguistic analysis of AR and sA-R, see sections 7.2 and 7.3 below.
“Gebrauchspräferenzen” meint die sprachliche Wahl aus einer Menge gegebener Alternativen für bestimmte Ziele und Zwecke. “Stil” ist somit Handlungszielen und -zwecken untergeordnet und teils strategischer, teils habitueller Natur.14
“‘Preferences of use’ means the linguistic choices among a variety of options to achieve certain goals and means. ‘Style’ is thus subordinate to pragmatic factors and goals, and it is partly of strategic, partly of habitual nature.”
Style as a strategic and proactive measurement ensures the efficiency of the “magic” formula (see section 1 above). It is noteworthy that shortenings, ellipses, and fragmentary utterances are characteristics of dramatic dialogue as well.15
See, for example, Betten 1985:269: “Bei der Analyse moderner Dramen-dialoge erweisen sich Verkürzungen und Wiederholungen als besonders beliebte und häufig gewählte Stilmittel.” [When analyzing modern dramatic dialogue, shortenings and contractions reveal themselves as extremely popular and frequent stylistic devices.]
In a comparative study, Fraenkel (1939:36) observed similar elisions, ellipses, and haplologies in Baltic curses: “Charakteristischerweise sind diese kürzeren Formantien besonders in Flüchen gebräuchlich” [Typi-cally, these shorter formations are common in curses.]16
With regard to runono, Fraenkel (1939:36–37) mentions close parallels of Indo-European invocations: “[Baltic] boža łośc hat Haplologie an der Wortgrenze erlitten wie ved. ápāv
k(ta) táma
, griech.
, etc.” [Baltic boža łośc underwent haplology at the word edge as Vedic ápāv
k(ta) táma
, Greek
, etc.]
In conclusion, Stentoften's fragmentary-elliptical style is more effective in a dramatic sense than Björketorp's more explicit wording. This reassessment is supported by the fact that the Björketorp text represents linguistically younger, syncopated forms involving umlaut and breaking, for example, bArutR, fAlAhAk,hAerAmAlAusR, as opposed to Stentoften's unsyncopated and hence older forms bAriutiþ, felAhekA, herAmAlAsAR (on the innovative Björketorp forms AR and sA-R, see sections 7.2–7.3 below). Already Jacobsen (1935:31–32, 42; my translation) argued that the formulas Stentoften and Björketorp represented two independent versions of an older source:
In such a fortunate situation, where two independent versions of an older text are preserved, it is unacceptable to rely exclusively on one of the two. […] Both the Stenthoften stone and the Björketorp stone are preserved in a complete form, such that all the inconsistencies of their sound representations are explicable in terms of the on-going struggle between actual speech and traditional writing, and that both inscriptions provide a natural and adequate reflection of the language of their time.17
“Det går dog visst ikke an, når man er i det heldige tilfælde at have overlevert to av hinanden uavhængige bearbejdelser af en ældre text, da i sin rekonstruktion kun at tage hensyn til den ene.” […] “Såvel Stentoften-stenen som Björketorp-stenen er bevaret i fullstændig form, at de inkonsekvenser i lydformer, der findes, alle lader sig forklare som den særlig i overgangstiden herskende kamp mellom levende tale og traditionel skrift, og at begge indskrifter giver et sandt og naturlig udtryk for deres tids sprogforhold.”
Summing up, there are strong indications of oral speech features in the Blekinge inscriptions: first, the accumulation of epenthetic vowels, second, the consistent representation of schwa values (reflexes of original theme vowels), and third, the condensed, elliptical style of the Stentoften curse in relation to the modernized Björketorp version.
6. Runic Sandhi Phenomena.
6.1. A Preliminary Remark.
The stylistic assessment of sentence sandhi in this section is solely concerned with consonantal phenomena. The reason for this restriction is that vowel sandhi in runic epigraphy is extremely difficult to evaluate since the double reading of single vowel runes is signaled by a space in a number of inscriptions, particular on bracteates, as in Darum bracteate (V)–C, Jutland (IK 43 = KJ 104; ca 450–550 C.E.) niujil alu → nijil(a) alu, or Ølst bracteate–C, Jutland (IK 135 = KJ 123; ca 450–550 C.E.) hag alu → hag(a) alu.18
On niujil(a), see Peterson 2004:14 and Nedoma 2005:171. Note that IK refers to Axboe, Düwel, Hauck, and von Padberg 1985–1989.
For this epigraphic rule, see Nedoma 2005:181–182, note 54: “Doppelte Geltung einer Rune über eine (durch Spatium markierte) Wortgrenze hinweg tritt in den älteren Runeninschriften ferner etwa in vor-a[lt]s[sächsisch] ks
mella lguskaþi = skamella (a)lgº Wremen, 425–450 (Düwel 1994:14–15) entgegen.” [The single representation of two runes across a word boundary is also found in pre-Old Saxon ks
mella lguskaþi = skamella (a)lgº Wremen, 425–450.]
6.2. Sandhi in Viking Age Inscriptions.
In a stimulating paper on Runic orthography, Lagman (1989:29–35) explained irregular (that is, non-standard) spellings in Swedish Viking Age inscriptions in terms of factors that have little to do with proper writing errors. In his wake, several authors (see, for example, Williams 1990, 1994, Meijer 1995, and Larsson 2002:14) stressed that runic orthography is in general remarkably correct, and that runographers base their spellings on sound analysis.
One of the deviations noticed by Lagman (1989) concerns the omission of runes in external sandhi position, for example, honsalukuþs for hans sálu ok guðs ‘his and god’s soul', or þuri :risþi for þóriR ræisti ‘þoriR raised’ (for parallel evidence from Runic Danish, see Makaev 1996:67–68). In Lagman's (1989:31) view, this san-dhi phenomenon constitutes a normative rule of runic writing:
A rune is left out in the external sandhi, a phenomenon which must be considered as a norm. [En runa sätts ut vid yttre sandhi, en företeelse som också betraktas som norm.]
-] from /-s#s-/, but also assimilations across word boundaries, such as k = [-k
-] from /-k#g-/, and r = [-R
-] from /-R#r-/, or u = [-
] from /-u#
-/. In a framework of fast-speech phonology, these instances presuppose the loss of juncture due to allegro rules. Following Dressler (1975b:19), “in allegro […] pauses vanish, processes extend over boundaries (cp. assimilations and migratory processes), that is, the units of programming are longer.” Let us now revert to the older runic inscriptions and look for related phenomena.
6.3. Sandhi in Transitional Runic.
In Transitional Runic, there is one striking example with the use of a single graph for a double consonant across a word boundary. The sequence welAdudsA in the Stentoften inscription (most likely for +welAduds#sA; see below) provides a central case of this type of sandhi rule.20
On the reading wel Aduds [ss]A, see among others Jacobsen 1935:30–32, 58–60, Krause 1966:217–218, Nielsen 1968:32, 36–37, Boutkan 1995:36, Makaev 1996:122 (with older literature), and Williams 2001:510. See also Antonsen 1975:84, as opposed to his later works such as Antonsen 1978:287 and 2002:311.
As a matter of fact, two major attempts have been made to justify the single spelling s by positing an endingless form -dud. One of these is launched by Antonsen (1978, 2002), who after having published his rune grammar in 1975, posits an endingless dative -dud. His proposal relies on the consistency of spelling practices in Early Runic (Antonsen 1978:287, compare 2002:311):21
As Antonsen 1975:12 remarks, there are practically no exceptions to this spelling rule in the corpus of the Early Runic inscriptions of Scandinavia. But note the allegedly East Germanic inscription on the Pietroassa gold ring (Wallachia, Rumania, ca. 300–400 C.E.): gutanio wihailag, interpreted as +wih+hailag ‘sacrosanctum’, which according to Makaev 1996:122 is “in full agreement with the principles of runic graphics;” see also Antonsen 1975:74.
Since inscriptions in the older fuþark do not represent double consonants separated by open juncture with a single rune […], St. -dud in the sequence dudsA must represent an endingless dative (under the influence of the consonant stems; Antonsen 1969–1970:65–70), rather than a genitive singular +-duds, as maintained by N.Å. Nielsen (1968:32).
From a historical perspective, however, Antonsen's argument is inconclusive since the inherited relic group of consonantal stems assumed the influential u- and a-stem endings, not vice versa; see, for example, ON dative singular f
te belonging to the masculine u-stem fótr‘foot’. Similarly, the old consonantal stems maðr ‘man’, nagl ‘nail’, vetr‘winter’ keep the original nominative accusative plural +-iR (for example, umlauted menn, mennr, also meðr), but otherwise follow the a-stems, hence dative singular manne, nagle, vetre (see Noreen 1923:§414, Heusler 1932:§244). In view of the Old Norse evidence, an endingless dative form -dud for the original masculine u-stem +dauþu- ‘death’ runs into morphological problems, as we would expect a full form +-dude or +-dudi instead; see Stentoften's -dAude, as well as Björketorps uti (< +ūtē < PG +-ai).
Another diachronic escape route is followed by Syrett (1994), who takes his departure from a masculine n-stem welA-dud literally ‘the one who has an insidious death’, with the ending lost (see Björketorp's welA-dAude and bahuvrihis such as ON dauð-váni adjective ‘deathly ill, near to death’, literally ‘having the expectation of death’).22
Thus, formally, the compound wel Ad Aude in the Björketorp legend is interpreted as a bahuvrihi in the nominative singular: ‘he who suffers an insidious death’ (see Krause 1971:92). The second option, as favored by Antonsen 1975:88, is a tatpurusha in the dative singular ‘[condemned to] an insidious death’.
If we suggest […] that [welAdud ] accurately represents the endingless stage after apocope of the final +-a but prior to the addition of the ending proper to the ijan-stems, the later stage which may be attested in KJ 97 Björketorp dAude, the stone gives a rationale for the distribution of the two forms […] the Stentoften carver went rushing straight on through the sentence at this point.
In fact, there is no good reason to assume that an inscription captures this postulated intermediate stage of apocope just before the introduction of the new ending of the ijan-stems proper (ON type han-i ‘cock’ in analogy with hlenn-i ‘robber’ from PG +hlanþ-ij-a). In his in-depth discussion, Nedoma (2005:171–172, 182, note 57) rightly dismisses both Antonsen's and Syrett's analogy-based accounts. We are therefore left with the reading +welAduds#sA (see Williams 2001:510).
No phonological explanation for this orthographic simplification has been provided in the literature. From a stylistic perspective, the Stentoften curse formula has already been put on a par with invocations, forms of address, and dramatic dialogue, which reveal similar elisions and shortenings (see section 5 above). Returning to the sandhi phenomena in Viking Age inscriptions, as mentioned in section 6.2, it is now possible to identify an allegro rule of informal styles involving the loss of juncture.23
Note that the simplification of the final cluster [-ðs#] > [-ð(:)#] can be compared to certain German dialects (see Harnisch 1995, 1998). Two related phonological rules of Bavarian are mentioned by Harnisch (1995:84), which involve the assimilation of a dental fricative [s] to a preceeding palato-alveolar fricative [∫] (e.g., [vri∫+s] > [vri∫∫] ‘frisches’), and the shortening (degemination) of the assimilatory product ([vri∫∫] > [vri∫]). The Stentoften sequence wel Adudcan thus be accounted for by a corresponding rule of progressive assimilation of a labiodental and an alveolar fricative [-ð+s#] > [-ð:#] > [-ð#]. But as no reflexes of such a rule are found in later Old Norse dialects, it remains purely hypothetical. In view of the Viking Age evidence as presented in section 6.2, sandhi is crucial here because of the immediately following [s] in sA. Due to the loss of juncture, the difficult phonotactic sequence of three fricatives +[-ðs#s-] is simplified to [-ðs(:)-], hence the single spelling s in wel AdudsA.
Being aware of this sandhi rule in 4b, Makaev (1966:67–68) adduces further evidence from Viking Age runic inscriptions: Snoldelev stone (DR 298–302) kunvAltstAin for +kunvAlts stAin ‘Kunvald’s stone', and the inscription from Vester-Marie 2 (DR 440–441), uk[u]þ(s) for +uk guþ(s) ‘and god’s'.25
See section 6.2 above. In addition, Makaev (1996:67) mentions the older runic inscription of Möjbro: frawaradazanahahaislaginaz (resolved as +is slaginaz ‘is slain’). However, this reading is not imperative; see Antonsen's segmentation ana hahai slaginaz‘F. on the stallion is slain’, with the auxiliary verb omitted; see Antonsen 1975:33–34. As Seebold (1994) remarks, the expected shape of the form in Early Runic should be istrather than is; see ist in the Vetteland legend (KJ 60; around 400 C.E.). Note that DR (followed by column) refers to Jakobsen and Moltke 1941–1942.
Recall that the sequence welAdudsA receives an exceptional status in the older runic corpus as it provides the earliest evidence for this allegro rule of connected speech. There are several instances of double spellings in all periods, including the end of the older runic period, as shown in 5.
The lento realization of the Rävsal inscription in 5a is confirmed by the graphic use of a word divider between two homorganic consonants [s#s]: a single dot marks the word seam between hAriwulfs and stAinAR. This is a conscious reflection of the word boundary between [-s#s-] that is based on lento speech. Similarly, the Eggja legend in 5b shows a double spelling in the sequence nissolusot for nīs sōlu sōtt, or further divided into ni (i)s sōlu sōtt.
To sum up, the Blekinge inscriptions receive special status within the corpus of the older runic inscriptions as they reflect actual speech phenomena and even a fast-speech rule with regard to sentence sandhi. Such sporadic reflections of phenomena of connected speech become more frequent in the younger runic inscriptions as an indication of more spontaneity and informality of runic writing in certain text types (see also Spurkland 2005). This substantiates the claim made in section 1 that due to the bifurcation of runic writing, a stylistic register responds to the characteristics of different types of runic texts. Clear evidence is found already in Transitional Runic.
7. Allegro Forms as Early Markers of Linguistic Change.
7.1. Preliminary Remarks.
Fast-speech phonology emphasizes that auxiliary verbs undergo phonological change earlier (also in more formal styles) than full verbs. More generally, function words are prone to linguistic change earlier than the class of content words (for example, Dressler 1974:26, 1975a:§5.1 and Harris and Ramat 1987 on grammaticalization processes in general). In what follows, it is claimed that allegro forms are present in the epigraphic sources of the transitional period. There are several instances of function words foreshadowing innovative developments such as rhotacism, in particular the auxiliary verb +is, ist and the preposition +aftr- in the Blekinge group around 600–650 C.E.
What characterizes allegro forms in terms of fast-speech phonology is that they are subject to fewer restrictions than the corresponding lento forms, and thus tend to be innovative. Dressler (1973, 1975a) points out that within a given language, processes often operate in allegro (casual speech) as “major rules” and in lento (formal style and citation form of speech) as “minor rules” (see section 1.2 above and Hinskens 1998:168–169):
C'est un fait général que, souvent les mêmes processus phonologiques se présentent comme des règles généralisées (règles dites “majeures”) dans les styles négligents et comme règles restreintes et exceptionnelles (règles dites “mineures”) dans les styles formels d'une même langue. C.-à-d. la forme et les effets de la règle y sont plus limités et, souvent, elle n'est appliquée qu'à des mots isolés” (Dressler 1973:133).26
“It is a general fact that the same phonological processes often constitute generalized rules (so-called ‘major’ rules) in casual styles, but restricted and exceptional rules (so-called ‘minor’ rules) in formal styles of the same language. This means that the form and the effects of the rule are more restricted here, and often, the rule is applied to isolated, frequent words only.”
The identification of allegro forms in Transitional Runic is apt to explain why there are several early (sporadic) indications of sound changes that otherwise are evidenced considerably later in runic epigraphy (see sections 7.2–7.4 below). The notion of allegro forms is corroborated by the evaluation of oral speech features as presented in sections 3–6. The stylistic assessment presented in section 5 is further confirmed by the sequences utiAR and sAR, which are discussed in detail below.
7.2. The Auxiliary Verb AR versus is.
Among early Nordic innovations is the coalescence of the reflex of PG +-z with r in unstressed position. Variation in the paradigm of the auxiliary verb, such as vas/vaR ‘was’ and es/eR ‘is’ is characteristic of the Viking Age period and later Old Norse.27
See Stroh-Wollin 1997, who invokes the notion of regional-chronological variation.
Obviously, most researchers identify AR (in the runic sequence utiAR) as a 3rd person singular present indicative auxiliary verb reflecting early North Germanic is, ist.28
See Krause 1966:216, Antonsen 1975:86, Schulte 1998:135–139, 2005a:27, 33–34, and 2005b:175–177 (with further references). See also Antonsen 2002:310: “Expected is +iR (< PG +ez-i, PIE +es-y), the 2nd person singular form substituted for the 3rd person singular ‘is’.”
Die Form aR (bzw. æR) ist sonst erst in bedeutend späteren Inschriften bezeugt. Noch die Inschriften auf dem Stein von Eggja […] und auf der Spange von Strand […] vor allem auch die Inschrift von Stentoften, haben die Form i(s); trotzdem erscheint eine Form mit dem später allgemein durchgedrungenen grammatischen Wechsel auch schon im 7. Jhd. wohl möglich.29
“The form aR (or æR) is otherwise only attested much later in runic inscriptions. Even the inscriptions on the Eggja-stone […] and the fibula from Strand […] and above all the Stentoften inscription exhibit the form i(s); nevertheless, a form that has undergone grammatical change seems possible already in the seventh century.”
Antonsen, too, reckons with an innovative form. In his view, AR represents /æR/, which means that the form has undergone the East Nordic development +ai > æ due to monophthongization and lowering (Antonsen 1978:278, 2002:310). In discussing this line of reasoning, Syrett (1994:44, note 15) again mentions chronological problems:
There seems something at least to be said for the position in Antonsen 1978:287 that East Norse confusion of accented +eand +æ has already begun to show itself in this form, although it must presumably be understood as a purely local innovation ahead of its time given that later runic evidence appears to continue to distinguish these two phonemes into the Viking Age, see Noreen 1904:§113.
To sum up, whatever explanation is favored, the Björketorp form AR (identified as a rhotacistic variant of the auxiliary verb +is under weak sentence stress) poses severe problems for chronology. In my view, it represents an allegro form, which is entirely consistent with the stylistic assessment of the curse formula where it occurs (see section 5).
7.3. The Clitic Relative Pronoun -R (sA-R).
The form sAR in the Björketorp curse also shows early rhotacism, thus supporting the analysis presented in section 7.2 above. The runic sequence sAR contains the relative pronoun (the later relative particle) affixed to the demonstrative base sa-: sAR [sa:ř] = PG +sa(-ez) < IE +so(-es) (compare ON sá er, sá es ‘he who’).30
See Noreen 1923:§469, note 1, Lindblad 1943:151–152, Rosenfeld 1955:103, Antonsen 2002:311, and Schulte 2005a:32–34.
7.4. Early Rhotacism after Apical: The Form AfatR.
Under this approach, the preposition AfatR in the Istaby inscription (KJ 98; 600–650 C.E.) with R instead of expected r provides further sporadic evidence for rhotacism, here after apical /t/. The linguistically older form after(PG +aftr-) in the Tune inscription (KJ 72; Norway, ca. 400 C.E.) has etymologically based r. Apart from the isolated instance in the Blekinge group, the confusion of original +z and r after apicals is evidenced only considerably later, as in raknhiltr sustiR ‘Ragnhild’s sister' on the Tryggvælde stone (around 900 C.E.).31
See Nielsen 2000:99, 130 and Antonsen 2002:306.
See Stoklund 1996:204–206. Nielsen (2000:258) states: “But on the Ribe skull fragment -z has replaced -r in uiþz ‘against’ (compare: Goth[ic] wiþra), although this may well be the only inaccurate use of -z (or -r) in this particular inscription.”
Early on runologists noticed inverted spellings with R for r after apicodental obstruents. Makaev (1996:100) summarizes the runic evidence as follows:
Of note is the use of the rune R instead of r after t in an inscription from the transitional period. The observed alternation of R and r, which first occurred in a position after dentals and interdentals (see Noreen 1923:192), is typical for runic inscriptions of the Viking Age (see DRI:1003–1005), that is, for Period 2 in Jacobsen-Moltke's terminology. It is possible that R instead of r was caused by its position after a weakly stressed vowel (see DRI: 970).
Unless we allow for an accidental orthographic error, the Istaby form AfatR (circa 650 C.E.) stands as an early testimony of the phonemic merger of (unstressed) z and r, once again in the Blekinge group. However, contrary to Grønvik (1981:42), there is no proof of rhotacism being complete in Transitional Runic around 600–650 C.E. (see also Nielsen 2000:258). This makes it clear that function words of purely grammatical status deserve special attention in tracing allegro forms.
7.5. Lenition of /θ/ to [ð].
The weakening of voiceless /θ/ to voiced [ð] in intervocalic position is also witnessed in the Blekinge inscriptions. The innovative form welAdAude in the Björketorp curse formula exhibits d for an originally voiceless fricative /θ/ (see, for example, OE dēaþ, OHG tōd ‘death’ versus the (vernerized) final obstruent of OE dēad, OHG tōt ‘dead’).33
The precursor of -dAude is PG +-dauθ- with expected þ for /θ/ in Early Runic writing; see PG +dauθan-, Gothic dauþus ‘death’. See Magnússon 1995:108, under dauði.
The split of the [đ] allophone from Gmc. +/d/ and merger with /θ/ may thus not have been completed until the eighth century when forms such as uiþz ‘against’ (cf. ON víðr; OE wiðer) and uþin Óðinn (cf. OE Wōden) carved on the Ribe skull fragment testify to the restructuring.
It is noteworthy that the voice distinction between /θ/ and /ð/ is still reflected in the forms -moþA (PG +mōþa-) versus made (PG +-ið
) in the Eggja inscription, which is dated to 700 C.E. on archeological grounds (see Nielsen 2000:127). We are therefore left with the fact that a series of innovative dialectal features, which are otherwise mirrored in runic writing considerably later, show up in the Blekinge group in the early seventh century.
8. Conclusion.
The goal of this paper was to demonstrate that certain runic texts, in particular the Blekinge curse formula, display linguistic features of informal styles as well as allegro phenomena including elisions and syntactic ellipses. It follows that the runological pursuit of a strict standard inevitably leads to the assessment of “contradictory linguistic forms” (Düwel 2004:136–137). On closer inspection, however, the general axiom that a rigid lento norm is a regular basis for runic epigraphy can be called into question. Runic inscriptions exhibit features and elements of particular styles, and thus testify to stylistic variation and stratification as found in non-runic texts (Biber 1988, 1995). As with other written documents, linguistic variables intersect with stylistic ones, depending on text type, content, language user, addressee, and other factors.34
On the variety of runic genres in Norway through the Middle Ages, see Spurkland 2005.
In fact, the style of the older runic inscriptions is predominantly linked to high prestige and status; see, for example, the runemaster formula of the ek erilaz-type (Järsberg stone, KJ 70; Sweden, ca. 450 C.E.), or the standard example of the golden horn B from Gallehus (KJ 43; Denmark, ca. 400–450 C.E.), ek hlewagastiz ¦ holtijaz ¦ horna ¦ tawido ¦ ‘I, Hlewagastiz, son of Holt, made the horn’. As noted in 6.3, the use of word dividers (four vertical dots represented here as ¦) is a noteworthy indication of the lento norm, which this carefully wrought inscription is based on.
Nonetheless, sporadic influence of the pronunciation on spelling in the case of Myklebostad asu-gasdiz ‘Ansu-gasti-’ (circa 400 C.E.) with d for expected t is almost indisputable.35
The general question whether the non-representation of nasals before obstruents in runic writing is based on phonetic judgment has been deliberately excluded from this study; see, for example, Tørvika stone A (KJ 91; Norway; ca. 400–450 C.E.) lada-warijaz (OHG Lantwari) (also English children's spellings such as bup ‘bump’, tet ‘tent’, thik ‘think’, etc.; Miller 1994:97). Although this indeed is plausible, it is difficult to provide conclusive evidence that would rule out the influence of Mediterranean writing principles. See Derolez 1998a:20–21, with reference to Morris 1988:127: “Among orthographic practices the non-representation of nasals, occurring both in environments where the nasal was (later) lost, for example asu- (OE ōs, ON áss), and in such as retained the nasal, for example, lada- (OE ON land) suggests that nasals were still present, and that the practice was borrowed. If so, ‘then Greek would be the most likely candidate for the lender (p. 127)’.”
See Antonsen 1975:14, 2002:297–314, Makaev 1996:51–52.
Finally, the elitist and marked status of runic writing has partly been overrated (see, for example, Beck 2001:8–18). Despite its exclusiveness, even early runic script cannot be disconnected from actual speech because—like any alphabetic writing system—it is intricately linked with the language that it represents (Feldbusch 1985). Thus, it is a generally acknowledged fact that speech exerts an influence on writing (see Coulmas 2003:107–108). In this regard, I disagree with Braunmüller (forthcoming), who in principal rules out influences of actual speech on the older runic inscriptions. In this paper, speech-related elements of runic writing have been demonstrated by means of the (even consistent) representation of epenthetic vowels and schwa values, and the stylistic condensations of certain runic text types (“magic” formulas) in the transitional period. The runic sandhi laws of transitional and Viking Age inscriptions lend further support to this claim.
Finally, the elitist and marked status of runic writing has partly been overrated (see, for example, Beck 2001:8–18). Despite its exclusiveness, even early runic script cannot be disconnected from actual speech because—like any alphabetic writing system—it is intricately linked with the language that it represents (Feldbusch 1985). Thus, it is a generally acknowledged fact that speech exerts an influence on writing (see Coulmas 2003:107–108). In this regard, I disagree with Braunmüller (forthcoming), who in principal rules out influences of actual speech on the older runic inscriptions. In this paper, speech-related elements of runic writing have been demonstrated by means of the (even consistent) representation of epenthetic vowels and schwa values, and the stylistic condensations of certain runic text types (“magic” formulas) in the transitional period. The runic sandhi laws of transitional and Viking Age inscriptions lend further support to this claim.
