The carvers no one remembered

Gufi ræisti stæin þennsi æftiR Olaf, sun sinn, dræng harða goðan. Hann varð drepinn i Æistlandum. Havarðr(?) hiogg s[tæin].

Gufi raised this stone in memory of Ólafr, his son, a very good drengr. He was killed in Estonia. Hávarðr(?) cut the stone.

The stone stands at Frugården in Norra Åsarp in Västergötland (Vg 181). The last sentence is a carver's signature: Havarðr hiogg stein, Hávarðr cut the stone. Yet no carver is given for the stone in the Scandinavian Runic-text Data Base – nor in the newer edition, Runor 2020. It is the database that nearly all statistics on runestones rest on. Ours too.

Enlarge ⤢

Half the signatures are missing

In his dissertation Mästare och minnesmärken (2007), Magnus Källström went through every runic inscription in which someone says that he carved, cut or painted. We laid his list beside the database's carver field, stone by stone. We counted only runestones with what he calls a carver's signature of the more certain kinds – a sentence such as Hávarðr cut the stone beside the stone's other text.

Uppland is nearly complete: 2 of 120 signed stones lack a carver. That is where research on carvers has been done, with Œpir, Ásmundr and Balli. Elsewhere the picture is different. In Västergötland 4 of 5 are missing, in Östergötland 3 of 6, on Gotland 3 of 4. Víkingr and Grímulfr on Ög 8, Lófi on Ög 64 at Bjälbo, Hrómundr on Hs 14 in Hälsingland, Tryggvi on Vg 130 and Fundinn on Vg 180 – all wrote their names on the stone, and none of them is in the database's carver field.

This is no accident of single records. The database's carver data rest largely on research on the Mälaren region and have only partly been entered for the rest of the country. The consequence shows in every statistical study of carvers that rests on the database – ours included. In Västergötland 1 of 131 runestones is listed as signed there; according to Källström there are 5. Uppland still dominates, even with his figures – one in ten Uppland runestones is signed, against one in twenty-six in Västergötland – so that signing was most common in the Mälaren region is no illusion. But the database makes the difference larger than it was, and when we wrote earlier that one per cent of Västergötland's runestones have a carver, it was the database we measured, not what research knows.

Can a model find them anyway?

If the carvers are missing from the database, perhaps they show in the stones. A carver leaves traces in the text: how he spelled stæin and æftiR, which word dividers he used, which formula the stone opens with. Vitki looks for groups of stones without a known carver that hang together as tightly in those traits as a known carver's stones.

In Uppland and Södermanland, where the carvers are known, it works. With the carvers hidden, 85 per cent of a group were the same carver's, against 69 per cent if one simply takes as many neighbouring stones. But that is exactly where the method is needed least.

Outside the Mälaren region we could test it on what Sveriges runinskrifter says. In a few places the editors tie two or three stones to the same carver. Of nine such pairs where both stones are in our analysis, 6 land in the same group. That sounds good, but neighbouring stones at the same distance often land in the same group anyway – 4.8 pairs would do so from nearness alone. And the group around Kölaby, which holds Hávarðr's stone, also holds Vg 130, signed by another carver, Tryggvi.

Enlarge ⤢

As honestly as we can put it: the model finds writing milieus, not carvers. The groups share rock and unusual spellings more often than the stones around them – two stones in the Kölaby group write eftR for æftiR, three in a group around Konungsund write rasti for ræisti – and that is a sign of something real: a workshop, a school, a district's way of writing. But a group can hold several hands, and we cannot see where the line runs.

Three questions for runologists

Three groups are not described as one carver's work in Sveriges runinskrifter or by Källström. We do not call them new carvers. But we have gone through them once more: with a model that estimates how likely it is that a group with exactly these traits is one master's work, with the literature, and with the signed stones in the neighbourhood – do they resemble the group more than the area's other stones do? Here is what we know, and how each question can be tested.

Konungsund (Östergötland, 28 stones) – a master and his neighbours? The group is large. It holds 44 per cent of the area's contemporary stones, more than nearly any known carver in the Mälaren valley – there the median is 12 per cent. The stones share their rock more often than the area's do (92 per cent of the pairs against 77), nearly all are granite, and they share unusual spellings: three write rasti and two resti for ræisti, two tsin for stæin. Between the group's oldest and youngest style group lie 32.5 years, which fits within one carver's life. The model says the group is almost certainly not a single hand – 8 per cent – but probably a master with some others' stones mixed in: 76 per cent, just below the usual.

The literature offers two threads. Erik Brate assigned Ög 46 and Ög 233 to the same carver: Vébjôrn's two stones, one for each of two brothers. And Ihrfors held, according to Brate, that Ög 225 in the group was carved by the same hand as Ög 231, which stands outside it.

At Östra Stenby church, next to Ög 233, stands Ög 234: Karl ræist stæin þennsa æftiR Asmund, broður sinn. Karl carved – or raised – this stone in memory of Ásmundr, his brother. The same Karl appears on Ög 235 in the same churchyard. Whether ræist means carved or is a shortening of raised cannot be decided, writes Källström (2007, pp. 85–88). He places Karl's stones among the uncertain cases, but notes that for other reasons they are assumed to be carved by the same person.

We measured how like the group Karl's stone is in the traits of the hand. As a whole it is not particularly close: 21 per cent of the area's stones outside the group lie as close. But among the area's hundred-odd stones, Ög 233 is the fourth and Ög 46 the sixth most like Karl's stone. Our guess therefore becomes narrower than before: Karl may have carved Vébjôrn's two stones, but hardly the whole group. That is tested on the stones themselves – rune forms and grooves on Ög 46, 233, 234 and 235.

Enlarge ⤢

Vetlanda (Småland, 14 stones) – was Þorsteinn the carver? Fourteen stones around Vetlanda that resemble each other more than nearly any group in the calibration: the model gives 98 per cent that they are mostly one master's work. But nothing else binds them – neither rock, names nor unusual spellings – and only one of the stones has a style group, so their time cannot be judged. Two of them concern the same man, Vrái: on Sm 76 at Komstad he is called Earl Hákon's marshal, and Sm 77 at Sävsjö he raised himself in memory of his brother Gunni, who died in England. At least two of the stones thus belong to an earl's man's circle. On Sm 111 at Fageräng the text ends in a damaged sequence of runes that Ragnar Kinander read as a prayer and later readers as a signature: aku- hio[gg], someone cut – perhaps an Ágautr (Källström 2007, pp. 208, 342).

What is new is a stone outside the group. Sm 96 at Brobyholm in Lannaskede, under twenty kilometres to the west, begins þurstin : rist stin: Þorsteinn raised – or carved – the stone in memory of Guði, his son, and he made a bridge. Kinander read raised, Lennart Peterson (1987) carved, and Källström leaves the question open. We measured the stone's likeness to the group. None of the 16 other stones in the area lies as close to the Vetlanda group as Þorsteinn's does, and his five closest stones in the whole area all belong to it – among them Vrái's stone and the Fageräng stone with the possible signature. It is the strongest result in this review.

The weaknesses are three. The area is small, so there are few stones to compare with. Sm 96 was not included in the group, because no single pair reached the model's threshold – it resembles the group as a whole, not any one stone especially closely. And Rundata has three readings of the text. The question to runologists is sharp all the same: did Þorsteinn of Lannaskede carve the Vetlanda stones? The rune forms on Sm 96 and Sm 111 are the first thing to compare.

Husaby (Västergötland, 6 stones) – the formula from Kinnekulle. Six stones on and around Kinnekulle, with no remark in the literature about the carver. The model gives 88 per cent for one master and 62 per cent that all six are one and the same hand – the highest of the three. The texts are short and almost alike. Five of six read X satti stæin þannsi æftiR Y – X set this stone in memory of Y – against 7 of 52 stones in the area. That is probably exactly what the model grouped them on, so it is a hallmark, not independent evidence.

Two of the stones carry images: a ship on Vg 51 at Husaby church, and on Vg 56 at Källby ås a standing figure with an animal head or mask and a snake belt round the waist (Stern 2013, pp. 34, 56). The nearest signed stone, Vg 59 at Norra Härene, where Hjalmr and Hjalli cut the runes, resembles the group no more than one stone in three in the area does. Hjalmr and Hjalli are thus probably not the Husaby carvers. The question becomes: is satti stæin þannsi one carver's formula or the district's? The images and rune forms on Vg 51 and Vg 56 can decide it.

How sure we are

Sources

• Scandinavian Runic-text Data Base (Rundata 3.1), Department of Scandinavian Languages, Uppsala University. Open Database License. Runor 2020, Swedish National Heritage Board.
• Källström, Magnus 2007. Mästare och minnesmärken. Studier kring vikingatida runristare och skriftmiljöer i Norden. Stockholm University. Appendix 1, pp. 333 ff.; pp. 85–88, 208, 342. Also Peterson (1987) on rist on Sm 96 there.
• Jungner, Hugo & Svärdström, Elisabeth 1940–70. Västergötlands runinskrifter (Sveriges runinskrifter V), pp. 20, 25, 29, 250 f., 296, 323–331.
• Brate, Erik 1911–18. Östergötlands runinskrifter (Sveriges runinskrifter II), pp. 210, 217.
• Stern, Marjolein 2013. Runestone images and visual communication in Viking Age Scandinavia. PhD thesis, University of Nottingham, pp. 34, 56, 337.
• Kinander, Ragnar 1935–61. Smålands runinskrifter (Sveriges runinskrifter IV), pp. 225, 249.
• Vitki's count, method and the locked plans are open on GitHub.
• Photos: Wikimedia Commons, with photographer and licence in the captions.