Sons and brothers died outside Scandinavia more often than fathers and husbands – 10 against 3 percent

Runestones for men who died abroad are among the most studied. Mats G. Larsson (1990) went through the stones about journeys abroad and what they say about the society behind them. Judith Jesch (2001) examined the vocabulary of runic inscriptions and skaldic verse on journeys and retinues, among them the words drengr and þegn. Lauren Goetting (2006) argued that drengr was used chiefly of young men in a warrior retinue and þegn of older, established men, and Denis Sukhino-Khomenko (2024) followed þegn around the North Sea. None of them measured whether the man who died abroad was more often a young man in his family than an older one, with time, district and carver held constant.

The aim of this study is to test that, with the family place as the measure: a son or brother stands in a younger place in the family than a father or husband. That is not the same as age, but it is the nearest measure the inscriptions themselves give.

The hypothesis, the model and the rules of interpretation were written down in an analysis plan on 28 October 2026, before any model had been run. The plan concerned a study of who raised the runestones and for whom, with six tests; this is test 5. The plan is time-stamped in the project's version control and has the SHA-256 checksum ea26a388379e29eb… (see Data and code).

The test reads in the plan: the outcome is that the stone says the dead man died outside Scandinavia, among stones where the message can be read; the dead man is young in the family when he is a child or sibling of the sponsor and neither parent nor spouse, and older when he is a parent or spouse and neither child nor sibling; only stones on which all the dead are men; the model is outcome ~ young + region + dating period + carver as a random effect. The six tests are corrected together by Holm's method. A test confirms when Holm p is below 0.05 and the direction is the same with p below 0.05 in each of the material's two halves; a pattern to test has Holm p below 0.05 but does not hold in both halves.

Before the plan was locked no count of death abroad by family place had been made. For tests 3 and 4 on drengr and þegn, however, I had seen a rough count without a model: drengr on 8 percent of the stones for sons and brothers against 0.4 percent for fathers, and þegn the other way round. The plan says so. Tests 3 and 4 are therefore reported in section 5.5 as complementary, not as main results.

Sections 5.1–5.3 are the analyses the plan prescribes for test 5. Section 5.4 is descriptive counts on the same selection, made after the test. No other analyses are included.

• Source. The Scandinavian Runic-text Database (Rundata), Department of Scandinavian Languages, Uppsala University: transliteration, normalisation, English translation, province, parish, carver and style group for each inscription. The material was frozen for the analysis on 29 October 2026.
• The stones. Viking-Age runestones in the Swedish provinces: 2,320 stones. Runestones only; other objects with runes are not included. Skåne and Blekinge are counted as Denmark in Rundata and are not included.
• Selection. Stones whose text is preserved well enough for the dead man and his relation to the sponsor to be read, on which all the dead are men, and on which the dead man is either young or older in the family by the plan: 393 stones for sons and brothers and 552 for fathers and husbands, 945 in all.
• The family place. The relation is read from Rundata's English translation by rules (son, brother, father, husband and their forms). The rules were checked against hand coding on a new sample of 40 stones they had never seen: the relation was right on 38 of 40.
• The outcome is that the stone says the dead man died outside Scandinavia: a word for death or a manner of dying near a place outside Scandinavia (east, west, Greece, England, Serkland, Ingvar's expedition and others). Having merely travelled does not count, nor dying on a journey at home. The rule was checked against hand coding of 205 stones: the rule's yes was right 92 times of 92 and its no 113 times of 113.
• Region by today's provinces (Svealand, Götaland, Gotland, Norrland) and dating period by the stone's ornamental style (Gräslund 2006): before 1010, 1010–1040, 1040–1070, after 1070, and undated as a group of its own.
• The carver. The carver Rundata gives. Stones without a known carver are grouped by parish.

4.1 The model. Logistic regression with the R package mgcv (R 4.6.1; Wood 2017): the outcome against family place, region and dating period, with the carver as a random effect, so that a prolific carver's way of wording the stones is not counted as a pattern. The shares are the model's marginal effects.

4.2 Correction and replication. Holm correction (Holm 1979) over the study's six tests. Internal replication: the material was divided once, for all the project's analyses, into two halves parish by parish, and the test must hold in each half on its own. The plan's robustness checks: only stones with at least ten legible words, the model without the carver effect, Uppland only, and only sons against only fathers.

4.3 Sensitivity analyses. The same fixed series as for every association the project reports: the raw model; adjusted for province and dating period; the carver as a random effect; the parish as a random effect; the stone's location as a spatial surface (only stones that stand in place); each half on its own; and without one province at a time (provinces with at least 30 stones). An analysis holds when p is below 0.05 and the direction is the same as in the adjusted model; without a province, significance is required only where at least two thirds of the stones remain. The p-values are also adjusted together for all 48 associations reported (Benjamini & Hochberg 1995); the adjusted value is called q. The model's assumptions are checked with simulated residuals (DHARMa; Hartig 2022). The errors in the outcome rule are tested by 200 refits in which each stone's ”true” outcome is drawn from the hand-coded sample's hit rates.

4.4 Deviations from the plan. None. The plan's check ”sex determined for every sponsor” applies only to the tests about the sponsors and is not made here.

5.1 The main test. Raw, the stone says the dead man died outside Scandinavia on 41 of 393 stones for sons and brothers (10.4 percent) and on 15 of 552 for fathers and husbands (2.7 percent). With time, region and carver constant the shares are 9.7 against 2.6 percent. The odds ratio is 4.10 (95 % interval 2.18–7.69), p < 0.001, Holm-corrected p < 0.001. In plain words: on a stone for a son or brother it is three to four times as common that the dead man died abroad as on a stone for a father or husband. Half A: 481 stones, odds ratio 3.03 (1.25–7.35), p = 0.01. Half B: 464 stones, odds ratio 5.82 (2.30–14.71), p < 0.001. The test confirms by the plan's rule.

5.2 The plan's robustness checks.

• Only stones with at least ten legible words: 661 stones, odds ratio 4.80, p < 0.001.
• Without the carver effect: odds ratio 4.02, p < 0.001.
• Uppland only: 489 stones, odds ratio 4.93, p = 0.003.
• Only sons against only fathers: 653 stones, odds ratio 5.26, p < 0.001. Raw 26 of 230 against 13 of 496.

5.3 The sensitivity analyses (945 stones, 56 with the outcome). All 8 hold:

• Raw model: odds ratio 4.17 (2.27–7.65), p < 0.001.
• Adjusted for province and dating period: 3.90 (2.09–7.27), p < 0.001.
• Carver as a random effect (394 carvers and parish groups): 4.41 (2.08–9.34), p < 0.001.
• Parish as a random effect (399 parishes): 4.04 (1.98–8.22), p < 0.001.
• The stone's location as a spatial surface (567 stones that stand in place): 2.56 (1.19–5.48), p = 0.02. The difference is smaller here, but it remains when the district is held constant.
• Half A (476 stones): 3.19, p = 0.01. Half B (469 stones): 5.53, p < 0.001.
• Without one province at a time (5 provinces tested): the direction holds without each of them and is significant everywhere. Weakest without Uppland, when 456 stones remain: odds ratio 3.64 (1.63–8.11), p = 0.002.
• The difference in percentage points in the adjusted model: 9.9 against 2.9 percent, 7.0 points (probably between 3.8 and 10.2).
• The model's assumptions: residual distribution p = 0.56, remaining spatial pattern p = 0.17, but overdispersion p = 0.032: the model does not capture all the variation between the stones, so the intervals are probably somewhat too narrow (see 6).
• Errors in the outcome rule: the association holds in 100 percent of the 200 refits; median odds ratio 3.65, range 2.37–4.17.
• Adjusted for all 48 associations the project reports: q < 0.001. The association is one of 15 that hold in every sensitivity analysis.

5.4 Who they were (descriptive, same selection). Of the 41 sons and brothers who died outside Scandinavia, 26 were sons and 15 brothers. On 13 of the stones a woman stands among the sponsors, and all 13 are mothers commemorating their sons. Of all 69 stones for men who died outside Scandinavia, whatever their relation to the sponsor, only 4 were raised by a wife. The difference is present in both large regions: in Svealand 26 of 261 for sons and brothers against 12 of 423 for fathers and husbands, in Götaland 15 of 124 against 3 of 121.

5.5 The words for the dead (the plan's tests 3 and 4). drengr is said of the dead man on 27 of 393 stones for sons and brothers against 4 of 552 for fathers and husbands: odds ratio 8.7, Holm-corrected p < 0.001, but too few cases in half B to test there. þegn stands on 5 of 393 against 26 of 552: odds ratio 0.22, Holm-corrected p = 0.01, but not significant in either half. Both are patterns to test by the plan's rule, and for both a rough count was seen before locking.

What the finding says. The men whose runestones say they died in the east or the west were more often sons and brothers than fathers and husbands. The difference is large, three to four times, and it remains with time, province and carver constant, in each half, in every province and with the district held constant. Only four of the 69 stones for men who died abroad were raised by a wife. The journeys seem to have belonged to a stage of life: the time before the man had a family of his own to raise the stone for him.

Against earlier research. The result gives a measure to the reading of the words for the dead that Goetting (2006) argued for: drengr for young men, þegn for older ones. Here both words follow the family place in the same direction as death abroad, but only as patterns to test. The finding on death abroad rests not on the words but on the relation, and so it is independent support for the journeys abroad having been chiefly a matter for young men (cf. Jesch 2001; Larsson 1990).

Weaknesses, in order of weight.

1. Family place is not age. A son may be middle-aged, a brother may be older than the sponsor. The measure says that the dead man was not yet the one whom others commemorated as father or husband, not how old he was.
2. Who raises the stone and what is written go together. A father who died abroad may have been commemorated by sons who worded the stone differently, without a place for the death. The finding concerns what the stones say, and it may partly be a difference in how different sponsors phrased themselves.
3. Overdispersion. The model check shows more variation between the stones than the model allows for (p = 0.032). The odds ratio is hardly affected, but the intervals are probably somewhat too narrow. With 56 stones with the outcome they are already wide.
4. The relation is read by rules. The rules had 38 right of 40 on a new sample. Errors in the relation are not part of the refits, only errors in the outcome.

What would overturn the finding. A test in which the sponsor is held constant – for example only stones a parent raises for a son compared with stones a son raises for a parent – in which the difference disappeared, or a hand coding of the relation showing more errors among the stones about death abroad.

On the runestones for sons and brothers it says more often that the dead man died outside Scandinavia than on the stones for fathers and husbands, 9.7 against 2.6 percent with time, province and carver constant, and the difference withstands every sensitivity analysis. The journeys abroad belonged chiefly to the young men of the family. The next step is to hold the sponsor constant and to test the words for the dead on a larger material.