Raise It Yourself – or Have It Raised
We counted the most common words on Sweden's Viking-age runestones, province by province and decade by decade. The words stayed home: Västergötland raised its stones with its own hands, Uppland had others do the work, Södermanland set its marks at – not æftiR – the dead. Eleven words refuse to be national, and thirty change shape in the middle of the eleventh century.
Anyone who reads many runestones notices it eventually: the stones say nearly the same thing, but not with the same words. One man raised the stone after his father. Another had it raised – and that little had holds a whole division of labour: one who commissioned and paid, another who carved. We have now counted the choices of words across the whole Swedish corpus, and they turn out to be neither chance nor taste. The choice of words follows the provinces, and it follows time.
The material is the 1,848 Viking-age runestones in the Scandinavian Runic Text Database with at least three legible words, and the 75 most common words on them – names excluded, since names follow families rather than regions. Each word is counted once per stone: either it is there or it is not. Of those words, eleven carry a regional signal that survives every control we can subject it to, and thirty change clearly over time.
Two ways to raise a stone
Uppland and Västergötland did the same thing with different grammar. On the stones of Västergötland the simple verb dominates: N ræisti stæin – 'N raised the stone' – stands on 52 per cent of the stones, against 10 per cent in the rest of the country. In Uppland the same word form is nearly extinct, 3 per cent against 22 outside. The Upplanders said let ræisa – 'had raised' – instead, and accordingly ræisa stands on 43 per cent of Uppland's stones against 17 per cent elsewhere.
ræisti ('raised')
ræisa ('raise', after let)
The scale runs from 0 to 60 %. Source: Scandinavian Runic Text Database, 1,848 Viking-age runestones with at least three legible words; Vitki's R module Ordförrådet. The shares are observed, the tests adjusted for time and text length.
The difference is not merely stylistic. Whoever had a stone raised paid someone else for the work – the construction carries a hint of commissioner and craftsman, of a division of labour that seems to have been self-evident around Lake Mälaren and uncommon in the west. No one took that idea further than Jarlabanke of Täby, who did not even wait to die: "Jarlabanki had these stones raised for himself while he was still alive ... and alone he owned all of Täby." We measure the words, not the society behind them, but the words point.
Something being more common in one province can have other causes than the province. The Uppland stones are on average younger than Götaland's, and short texts hold fewer words than long ones. Every word is therefore counted in a model that holds time (the style-group dating) and text length constant, and that moreover treats the carver as a source of similarity of its own – otherwise a single prolific carver's habit could look like a whole province's. Finally, each province's parishes are split into two halves, and the signal must point the same way in both.
The word that draws a line around Södermanland
The memorial formula's little preposition – æftiR or at – is the sharpest boundary in the whole material. On the runestones of Södermanland, at stands on 55 per cent of the stones; in the rest of the country on 19. The mirror image: æftiR stands on 32 per cent of the Södermanland stones against 68 per cent outside. The same formula, the same meaning – but the people of Södermanland chose the short word, and the neighbouring provinces the long one. The word at passes all the controls with a wide margin: the probability of seeing so large a difference by chance, if the choice of words were really the same across the country, is less than one in a thousand trillion.
at
æftiR
The scale runs from 0 to 70 %. Source: the same material and model as above.
Västergötland's dræng and Uppland's souls
Two more words that refuse to leave home. On the stones of Västergötland the dead are praised as goðan dræng – 'a good young man': dræng stands on 12 per cent of the stones there against 1 per cent in the rest of the country, and goðan on 30 against 6. That fits what we have seen before: the formula language of Västergötland points south, towards the Danish area, where the drengs live in the stones too.
In Uppland it is instead the soul that stands out. The prayer's and – 'spirit, soul' – stands on 13 per cent of Uppland's stones against 5 per cent elsewhere. Uppland's many late, Christian stones pray for the soul more often – and since the model holds time constant, this is not merely an effect of Uppland carving longest. Jarlabanke again: the bridge he made fyr and sina, for his soul.
The words that went out of time
Thirty of the 75 words rise or fall clearly across the eleventh century. Together they show the book of formulas changing shape. The verb gærði – 'made' – stands on 10 per cent of the oldest stones and falls to zero on the youngest. Letu – 'had raised', several commissioners – goes the other way, from 5 per cent to 44 at its peak, and risti – 'carved', the carver's own verb in the signature – grows from nothing to 19 per cent as carvers towards the end sign their names ever more often. The deictic þenna – 'this stone' – is halved, from 29 per cent to 10.
gærði ('made')
letu ('had raised')
risti ('carved')
þenna ('this')
The scale runs from 0 to 50 %. Source: the 972 dated stones; the dating comes from the ornamental style groups (Gräslund), hence roughly 980–1130.
The dating deserves its own caveat: the years come from the ornamental style groups, not from dates on the stones, so the curves describe the style groups' time. But the directions hold when we adjust for province and text length, and they tell an intelligible story – the commissioner formula takes over, the carvers sign, and the old gærði dies with its generation.
Which provinces write most alike?
Laying the provinces' whole word profiles side by side – with correspondence analysis, the same method archaeologists use to order grave finds – Södermanland and Uppland end up closest to each other, with Östergötland just beside them. Västergötland stands on its own, pulled towards its ræisti, its dræng and its goðan. It is the same picture that the language traits gave in our earlier analyses, but now computed from something as simple as which words stand on the stones at all.
What does it mean that the words stayed home? Put cautiously: that the memorial formulas were not a national standard but local traditions, stable enough to survive adjustment for time, text length and individual carvers. What we cannot say is why – whether it is dialect in speech, formula habits in the workshops, or the taste of the commissioners. The words do not distinguish the three. But the boundaries they draw are real, and they can be tested further: every table in the analysis can be downloaded, and anyone who likes can begin by reading one stone from Södermanland and one from Västergötland side by side. Listen for the at – and for the dreng.
Data: Scandinavian Runic Text Database (Uppsala University). The analysis is free software in Vitki; every figure in this article is computed from the Ordförrådet module and locked against its results in the article build.