Runic and Skaldic Evidence of Palatal r in West Norse

Runic evidence shows that the phonemes r and palatal ʀ merged at an early date in West Norse. I argue here that skaldic poetry also comprises valid evidence of this merger and that there is no reas ...


Introduction
E arly Old Norse distinguished between r which descended from Indo-Euro pean r, and a phoneme conventionally represented as ʀ, ultimately de scended from Indo-European s voiced to z by Verner's law. The exact reali zation of ʀ is unknown but since it affected preceding vowels in the same way as i did (ʀ-mutation) it is often referred to as palatal r (see further Peterson 1983, Larsson 2002. The transition from z to ʀ can be traced through internal developments, in particular the effects of final devoicing (Thöny 2017). There is also useful comparative evidence in Finnish, where z in the oldest Proto-Norse loanwords occurs as h whereas ʀ in a younger stratum of loanwords appears as r (Schalin 2016, 253-55). The ʀ phoneme eventually merged with r in all the Nordic languages and is not represented in any surviving manuscript. It is, however, represented by a special symbol in runic writing, the ýr rune. From runic orthography it appears that ʀ merged with r early in Norway and its colonies, survived longer in Denmark, longer still in Sweden - and lasted longest in Gotland where it survived into the 1200s (Larsson 2002). We have a very substantial corpus of East Norse inscriptions which distinguish ʀ from r.
This article reviews the chronological evidence for the merger of ʀ and r in West Norse and its main contribution is found in the attention given to skaldic poetry from the Viking Age. While this evidence has been referred to by previous scholars, it has always been treated summarily and sometimes dismissively. I attempt to remedy this here.

Runic evidence
From Viking Age Norway we have approximately sixty runestones along with a handful of runic inscriptions on other types of artifacts. Most of these runestones are from the end of the period, dating to the 1000s. It is abundantly clear that the distinction between r and ʀ had been lost by that time - the ýr rune is either not used at all or repurposed to stand for y and infrequently also for other vowels.
The Norwegian evidence from the 800s and 900s is much more sparse, especially considering that not all inscriptions have data pertaining to the r/ʀ question and many are difficult to date. As far as I can tell, there are only two Norwegian runestones from the Viking Age where the ýr rune is used for ʀ, the Valby stone and the Bjørneby stone.
The Valby stone (N 140) is conventionally dated to the 800s (Spurkland 2005, 76 f.) and contains a clear instance of etymological ʀ in auarþʀ (Old Ice landic ávarðr or Hávarðr). The Bjørneby stone (N 2) is a more difficult case - the runes are very worn but there seems to be agreement that the in scrip tion contains the word stątʀ (Old Icelandic stendr). There is little solid evi dence to use in dating the stone: Sanness Johnsen (1968, 98) proposes "ca. 800 eller tidligere" while Birkmann (1995, 328) suggests the 900s.
Claims have been made about the presence of ʀ in two further West Norse inscriptions. One is the pail from the Oseberg ship (c. 835) which has carvings that have been read as asikriʀ (N 138; Old Icelandic á Sigríðr), although this is far from certain. Michael Barnes notes that "some see here little more than a collection of vertical scratches" (Barnes 2012, 89). I have not personally inspected the pail but I find it difficult to reconcile available photographs with the traditional reading.
Another putative ʀ is found in the Isle of Man inscription referred to as MM 113 Andreas IV, the last visible word of which has been read as aftiʀ. When Katherine Holman inspected this badly damaged inscription in 1993, she was unable to confirm the reading of a final short-twig ýr rune and suggested that the rune in question might be "the remains of a long-branch r-rune" (Holman 1996, 107 f.). She further noted a drawing which appeared to show the same inscription in a better state of preservation - complete with an r rune rather than an ʀ rune. More recently, Michael Barnes argues that the visible remnants are more consistent with ʀ than r (Barnes 2019, 106). Whatever the case may be, a number of Manx inscrip tions are contemporary with Andreas IV and none of these show the use of ʀ. It must be regarded as very doubtful that this inscription tells us any thing about the use of ʀ in West Norse; as Barnes points out "there are likely to have been groups of people from eastern Scandinavia on the island as well" (Barnes 2019, 77).
There is an inscription with ʀ from St Albans in England (E 13) but it has East Scandinavian characteristics and so does not concern us here ( Barnes and Page 2006, 328). The same can be said of the deer's antler from Dublin (IR 12). On the whole, we have no convincing evidence of the use of ʀ in West Norse inscriptions beyond the 800s.
We can now move ahead to consider the use of r in place of etymological ʀ. The oldest Norwegian manifestation of this is found on a bronze ladle (Trå, found in a grave archaeologically dated to the first half of the 900s. The inscriptions on the ladle contain four cases of r written for etymological ʀ, including the final consonant of the word uritar (Old Ice landic óréttar) written twice (see e.g. Sanness Johnsen 1968, 200 f.). As far as I can tell, this is the only Norwegian inscription from the early or mid-900s which bears on the question. The oldest runestones to show r for palatal ʀ are, at the oldest, from the late 900s. One example is the older inscription on the Alstad stone (N 61), which appears to be pre-Christian and contains the preposition ur (Old Icelandic úr/ór). A number of Norwegian inscriptions dating to c. 1000 or the early 1000s affirm the disappearance of palatal ʀ.  1968,. These contain words such as sunr, runar and biarnar (Old Icelandic sunr, rúnar, Bjarnar) which demonstrate the merger of ʀ and r. There is also one inscription in the Hebrides (Kilbar, Sc 8) of similar age which reads r * str (Old Icelandic reistr). Runologists continue to wrestle with the early dating of these inscrip tions, which is based on art-historical considerations (Barnes 2013;Barnes 2019, 57-63).
The oldest runic inscription from Greenland, the Narsaq inscription dated to c. 1000, GR 76 (Imer 2017, 251) has some evidence to consider. The words haitir mar (Old Icelandic heitir maer) show the change clearly. As far as I can tell, there are no relevant runic inscriptions from Iceland, the Faroes, Orkney or Shetland which are old enough to add greatly to the picture.
To summarize the runic evidence, we have reasonably strong indications from the mid-to late 900s that palatal ʀ had merged with r in West Norse. The Trå ladle may push the date back slightly further. The dating of the Valby and Björneby stones is quite uncertain and we must be careful to avoid circular argumentation: a part of the reason these inscriptions have been classified as early is the very fact that they have palatal ʀ. Both inscrip tions are short and lack firm external evidence for dating - I fail to see how a dating of either inscription as early as c. 800 or as late as c. 900 can be ruled out. With this uncertainty, the merger of ʀ and r - based on the runic evidence alone - can hardly be dated more precisely than to the 800s or the beginning of the 900s.

Skaldic poetry as a source
Medieval Icelandic sources have preserved a considerable amount of orally trans mitted poetry attributed to poets of the pre-literary age. This in cludes a substantial amount of poetry from the mid-900s onwards and a small amount of earlier material.
The dróttkvaett metre was the mainstay of skaldic poetry and the only metre which we need to consider here. In classical dróttkvaett, the pen ultimate syllable of each line rhymes with one of the syllables occur ring earlier in the line. In the odd lines the rhyme involves one or more con sonants; in the even lines, the vowels match as well. There are some variations on these patterns, especially in the most archaic poems, but generally speaking it is easy to identify skaldic rhyme. Here are two examples: mǫŕ valkastar bǫŕu (Haustlǫng 3.6, 1 SkP, 3: 435) Hugstóran biðk heyra (Vellekla 1.1, SkP, 1: 283) The line from Haustlǫng is an even line and its rhyme includes the vowels in mǫŕ : bǫŕu. The line from Vellekla is odd and the vowels are different, stóran : heyra. Both examples show rhyme between r and earlier ʀ, attesting to the merger of the phonemes. The words heyra (Gothic hausjan) and mǫ-r (the nominative suffix from ʀ) derive from ʀ-forms while bǫŕu and stóran do not (as shown by the lack of ʀ-mutation and other evidence).
As discussed above, it is usually a relatively simple matter to demonstrate that a given skaldic corpus (such as a particular poem or the oeuvre of a particular poet) does not distinguish between r and ʀ. The value of this as linguistic evidence might be questioned in at least two ways, however.
One objection is that equivalence in skaldic rhyme is no guarantee of phonemic equivalence - the rhyme tolerated certain types of imper fections. The major types are as follows: 2 (a) Short and long consonants were frequently rhymed (e.g. t and tt). (b) Short and long vowels were occasionally rhymed (e.g. y and ý). (c) Oral and nasal vowels were treated as equivalent. (d) The phonemes a and ǫ were often rhymed. (e) The phonemes á and ǫ́ were often rhymed. (f) Velar k and g were treated as equivalent with palatal k and g respec tively.
Carl J. S. Marstrander (1915, 116 f.) and Börje Tjäder (1961, 11 f.) cited the imprecisions of skaldic rhyme as a reason to discount the skaldic evidence for r and ʀ. While some vigilance is certainly warranted, I think this is over cautious. For one reason, the known imprecisions of the rhyme affect mostly the vowels which the skalds may have regarded as less important. Skaldic rhyme has consonant rhyme in every line but vowel rhyme only in every second line.
Furthermore, none of the known imprecisions are analogous to a putative rule allowing equivalence between r and ʀ. Historical reasons can explain the equivalence of a and ǫ. The phoneme /ǫ/ emerged late and the poets may have imitated the works of earlier poets whose language did not distinguish between a and ǫ. The same applies to the palatal and velar stops but since this involves consonantal rhyme it may be worth investigating the matter in more detail.
The evidence of umlaut occurring before gi or ki shows that palatalization of g and k must have taken place at an early date. The e vocalism in degi from Proto-Norse *dagē must be due to the palatalization of g before front vowels since final ē in Proto-Norse did not occasion i umlaut (see e.g. Jón Axel Harðarson 2007, 79). It can then be assumed that g and k were pala talized in Old Norse before front vowels, as they still are in modern Ice landic. Yet the poets did not hesitate to rhyme velar and palatal stops, as in these examples (where underlining indicates the pertinent combinations: g or k with following i is palatal but velar with other following vowels): byrgis bǫðvar sorgar (Vellekla 4.3, SkP, 1: 287) jarls ríki framm slíku (Vellekla 12.8,SkP,1: 298) If velar and palatal stops could rhyme, it might be expected that r and pala tal ʀ could also rhyme with one another. I do not believe, however, that that conclusion necessarily follows. As with the case of a and ǫ, the velar and palatal stops arose historically from the same phonemes. This does not apply to the r/ʀ distinction since r and ʀ did not share the same origin. A poet who recognized these sounds as distinct units in his language would have no way of knowing that they would merge in the future.
Some scholars have attempted to explain rhyme between a and ǫ on a synchronic basis by invoking the structuralist idea of partially neutralized oppo sition (Hreinn Benediktsson 2002, 96). I have previously discussed the weak empirical basis of this idea and argued for other solutions (Haukur Þorgeirs son 2013, 16-19 and passim, see also Manaster-Ramer 1994). If we never theless briefly entertain the idea here, it might be suggested that certain phono logical developments created a special affinity between r and ʀ even prior to the merger of the two phonemes.
As an example, the inscription on the Glavendrup stone from the early 900s (DR 209) shows that the distinction between r and ʀ is present, with words like runaʀ and suniʀ. It also shows, however, that ʀ has developed into r after d in raknhiltr and has been assimilated to r in þur (Old Ice-landic Þórr). The question is whether developments of this sort would have been enough to create the impression among speakers that ʀ and r were inter change able sounds that could be rhymed with one another. I find this unlikely. Neither in Old Norse nor other languages does loss of the distinc tion between two phonemes in one or two environments normally lead to their rhyming together elsewhere. It could be noted that ʀ was also assimi lated to l, s and n (steinn < *stainʀ etc.) with no such consequences.
Ultimately, r and ʀ were different phonemes with different historical origins and different runic representations. There are no strong reasons to expect them to rhyme before their actual merger. Nonetheless in this matter, as in so many others, we must admit to some residual uncertainty.
Another possible objection to the use of skaldic evidence for linguistic developments is skepticism concerning the authenticity or reliability of the poetry which Icelandic written sources from the 1200s and 1300s attribute to Viking Age poets. Are the poems in question really works of the poets to whom they are attributed or are they later fabrications? Even if the attributions were made in good faith, was the oral tradition faithful enough to transmit appreciable amounts of poetry through 200-400 years in a form close enough to the originals to be useful for linguistic analysis? I believe the answer to both questions is a qualified yes.
Skaldic poetry attributed to pre-literary poets shows a wealth of archaic linguistic traits and a convincing chronology of metrical developments. This is emphasized in a detailed study of the question by Klaus Johan Myrvoll who notes (2014, 331 f.): Ein finn mynster som det vilde vera vanskelegt å gjera greida for um det var soleis at diktingi frå dei tidlege hundradåri hadde vore full av seinare fabri kasjonar. Det me oftast ser, er snarare ein jamn historisk vokster som er som ein skulde venta um storparten av det yverleverte materialet er autentisk.
Nevertheless, skaldicists generally believe that the individual stanzas (lausa vísur) attributed to the heroes of the Icelandic Family Sagas constitute less reliable material than the verses preserved in other sources. The poetry quoted in royal biographies and in essays on poetry is deemed to be more authentic. Fortunately, this includes the oldest material - the verses by Bragi Boddason and Þjóðólfr ór Hvini.
Survey of r/ʀ rhyme in Viking Age poetry As a basis for my survey I have taken the corpus of dróttkvaett poetry delin eated by Klaus Johan Myrvoll. Myrvoll includes all the poets who have at least eighty lines of verse and, for the oldest period, a selection of poets with a smaller preserved oeuvre (Myrvoll 2014, 35-37). He has further simpli fied matters by counting the number of rhymed lines, which is of relevance here (pp. 137 f.). Myrvoll's corpus extends to the 1200s but for our purposes here the Viking Age poets will suffice.
Within this corpus I have identified the following instances of rhyme between r and ʀ. I have underlined instances of r derived from ʀ.  (2) Pronouns (ʀ): There are thirty-seven cases involving pronominal forms, such as vér, mér, þeira and órar. Again the presence of ʀ is well attested through comparable runic and Gothic evidence. I have used Íslensk orðsifjabók by Ásgeir Blöndal Magnússon (1989) extensively as a source for the etymology of particular words and morphemes and have also consulted Altnordisches etymologisches Wörter buch (de Vries 1962) and Våre arveord (Bjorvand and Lindeman 2007) for difficult cases. In most words with r there is no controversy over whether descent from ʀ is involved but in a few cases there is no convincing etymology.
One example of a difficult word is keyra ('to whip, to drive') which is found in neither runic nor Gothic and has no compelling Indo-European cognates. Jan de Vries and Ásgeir Blöndal Magnússon entertain the idea of a derivation from Proto-Norse *kaurian, with a tenuous connection to a Serbo-Croatian word. Bjorvand and Lindeman (2000, 462) instead suggest a derivation from Proto-Norse *kauʀijan as a causative of kjósa, but this is far from semantically con vincing. The word occurs three times in rhyme in the corpus and since the etymology is uncertain I count none of these instances as rhyme between r and ʀ: Keyrum hross, svát heyri (Sigvatr, SkP, 1: 600) Keyrði Gefn ór garði (Óláfr inn helgi, SkP, 1: 528) Ofan keyrðum vér -orðum (Þjóðólfr Arnórsson, SkP, 2: 90) Another problematic word is eyrendi, which has plausible etymologies for both r and ʀ. There are doubtless further cases where another scholar might have included an example I exclude or vice versa.
The results are summarized in table 1, with the poets in chronological order.

Interpretation of the skaldic evidence
It is immediately apparent that rhyme between r and ʀ is widespread in the corpus, as noted already by Finnur Jónsson (1921, 262 f.). As we would have expected from the runic data, the poets of the mid-to late 900s uphold no distinction between the two sounds. Furthermore, Haustlǫng by Þjóðólfr ór Hvini has five examples of such rhyme. This is important evidence since its conventional dating of c. 900 makes it the second most archaic poem we have.
Intriguingly, the oldest extant poetry, by Bragi Boddason, contains no examples of rhyme between r and ʀ. Since Bragi's work is traditionally dated to the 800s it seems likely that this is no coincidence. Apart from Bragi's work, the corpus under investigation has 162 r/ʀ rhymes in 6,792 rhymed lines. If Bragi rhymed r with ʀ at this rate, the probability of obtaining at least one such rhyme in 101 rhymed lines can be calculated as follows: The result of 91% is below the conventional 95% threshold for statistical significance but this is not a study which can be redone with additional data to obtain more conclusive results. The skaldic and runic data that have come down to us are very limited and all scholars can do is attempt to make as much sense of them as possible.

Conclusion
The evidence reviewed in this article is summarized in table 2.
For the period of transition, both the runic and the skaldic evidence is quite sparse but nevertheless the two sources of data seem to be in harmony. The Trå ladle is archaeologically dated to the first half of the 900s and is the earliest runic evidence of the merger of ʀ and r. The poem Haust lǫng by Þjóðólfr ór Hvini is conventionally dated to "the late ninth or early tenth century" (SkP, 1: 3). These two sources clearly show the merger of the two phonemes which is confirmed by all the younger evidence.
The Bjørneby and Valby stones show the retention of ʀ. Both are difficult to date but can be tentatively assigned to the 800s or c. 900 at the latest. As for Bragi Boddason, it seems that "most scholars have accepted that Rdr