Children inherited parts of their parents' names – parent and child share a name element almost twice as often as chance gives

Germanic personal names are often compounds of two elements, and onomastics has long described how the elements passed on within the family, through variation (an element recurs, Holmger and Holmfast), alliteration (the same initial sound) and repetition of the whole name (Janzén 1947). Lena Peterson (2007) has listed the names of the runestones and their elements. Birgit Sawyer (2000) read the runestones as expressions of family and inheritance.

What is missing is a measure of how often the name elements actually pass on on the runestones, compared with what chance gives when one allows for some elements being common in a district, and whether the custom changed when Christianity came. The aim here is to measure that.

The hypotheses, measures and rules of interpretation were written down in an analysis plan on 29 October 2026, before any name pairs had been compared. The plan has the SHA-256 checksum 74eb51136233aa89… (see Data and code). Before locking only the number of relations was known and one example, U 135, where the son bears the father's name.

The plan tests five hypotheses with Holm correction: H1 that parent and child share name elements more often than chance gives, H2 that they alliterate more often, H3 that the excess is smaller on late stones, H4 that it is smaller on stones with Christian markers and H5 that siblings share name elements more often than chance gives. A hypothesis holds when Holm p is below 0.05 and the direction is the same with p below 0.05 in both halves.

The plan did not contain the project's fixed sensitivity analyses, which are built for associations between two traits on a stone. The corresponding steps for name pairs were therefore fixed in an addendum plan on 31 October 2026 (SHA-256 9d0247a8c21ba9b7…), before they were computed. One of the steps, without short elements, had been computed after the main run and was locked as a step in the addendum plan. A shuffle within the hundred (section 5.4) is exploratory: it was computed after the addendum plan, to test a weakness the plan did not address.

• The stones. Swedish Viking-Age runestones in the Scandinavian Runic-text Database (Rundata). Runestones only.
• The relations are read by rules from Rundata's normalisation and translation: father, mother, son, daughter, brother and sister. The rules were checked against hand coding on a new sample of 40 stones, where the relation was right on 38. Where a sponsor has the same name as the dead father, mother or son, the pair is also counted.
• The pairs. 1292 parent–child pairs and 1066 sibling pairs, each pair of names once per parish, so that a family with many stones, like Jarlabanke's, is not counted several times.
• The name elements. A name is divided into a first and a second element when both elements occur in at least three other names in the material; otherwise the whole name counts as one element. Two names share an element when they have an element in common or are the same name.
• Time by the stone's style group (Gräslund 2006): early stones RAK–Pr2, late Pr3–Pr5.

4.1 The measure. The share of pairs sharing a name element is compared with the share when the second name in each pair is shuffled at random among the pairs within the same province, 9,999 times. The shuffle keeps the province's stock of names, so that an element common in a district is not counted as inheritance. The test is one-sided.

4.2 The other hypotheses. H2 as H1 with the same initial sound as the measure. H3 and H4 compare the excess per pair, observed minus expected, between late and early stones and between stones with and without Christian markers respectively, two-sided.

4.3 Correction and replication. Holm correction (Holm 1979) over the five tests. Each hypothesis is also tested in each half on its own.

4.4 The sensitivity analyses under the addendum plan, for H1 and H5: the halves; without one province at a time (provinces with at least 30 pairs; significance required only where at least two thirds of the pairs remain); time held constant, with the shuffle within province and period; one pair per stone; without short elements and the data-driven junk elements -ni, -bbi and -einn; and only shared elements, without pairs where both bear the same whole name. A step holds when the share is greater than the expected one and p is below 0.05.

4.5 Deviations from the plan. None. The addendum run gives the same shares as the main run; the halves' p-values differ in the last decimal because the random draws come in a different order.

5.1 Parent and child (H1). 1292 pairs: 5.3 against 2.8 percent, p = 0.0001, Holm-corrected p = 0.0005. Half A: 4.9 against 2.3 percent, p = 0.0002. Half B: 5.8 against 3.4 percent, p = 0.0005. H1 holds. In plain words: about one parent–child pair in twenty shares a name element, against one in thirty-five if the names were distributed at random within the same province. The elements most often passed on are -björn, Tor-, -sten, -fast, Sig- and -mund, for example Holmger and Holmfast (U 289), Ingevald and Ingemund (U 296) and Gudmar and Gudbjörn (Sö 164). On 6 stones the son bears his father's whole name, as Östen after Östen (U 135).

5.2 Siblings, alliteration, time and Christianity (H5, H2, H3, H4). Siblings: 1066 pairs, 6.7 against 2.9 percent, p = 0.0001, Holm-corrected p = 0.0005, and in both halves (p = 0.0001 and p = 0.0001); H5 holds. Alliteration between parent and child: 14.9 against 12.5 percent, Holm-corrected p = 0.013, but only in half B; an indication. Over time: the excess is 3.1 percentage points on late stones and 2.6 on early ones, p = 0.89. Christian markers: 1.1 against 2.9 percentage points, p = 0.22. No difference is shown for time or Christianity. Mother and child: only 27 pairs, too few to say anything.

5.3 The sensitivity analyses. All hold, for both parent–child and siblings:

• Time held constant: parent–child 5.3 against 2.8 percent, p = 0.0001; siblings 6.7 against 3.3 percent, p = 0.0001.
• One pair per stone: parent–child (714 pairs) 5.6 against 2.8 percent, p = 0.0001; siblings (488 pairs) 7.2 against 2.5 percent, p = 0.0001.
• Without short elements: parent–child 4.5 against 2.5 percent, p = 0.0001; siblings 6.0 against 2.7 percent, p = 0.0001.
• Only shared elements, without the same whole name: parent–child 4.9 against 2.3 percent, p = 0.0001; siblings 6.7 against 2.3 percent, p = 0.0001.
• Without one province at a time: without Uppland 534 parent–child pairs remain, fewer than two thirds, and the direction holds (4.5 against 2.8 percent, p = 0.011); without each of the other provinces with at least 30 pairs p = 0.0001. The siblings likewise.

5.4 Shuffling within the hundred (exploratory). Families in the same district share a local stock of names, and shuffling within the province may therefore overstate the inheritance. If the names are shuffled only within the hundred, the expected share for parent–child becomes 3.3 percent against the observed 5.3, p = 0.0001, in 113 hundreds: the finding remains. For siblings the expected share becomes 5.6 percent against 6.7, p = 0.053: the excess shrinks and is no longer clear. That shuffle is however biased upwards for siblings: a stone with several siblings gives many pairs, and in a small hundred the random partner is often another sibling in the same family.

What the finding says. Children received elements of their parents' names more often than chance gives, also with the district's stock of names held constant, and regardless of time, province and how many names a stone has. But the custom concerned a minority: about one pair in twenty, against one in thirty-five expected. It was a possibility in name-giving, not a rule. That the excess is as large on late stones and on stones with Christian markers shows that the Nordic naming custom did not weaken in the families during the runestone period.

Against earlier research. The result gives a measure to the naming custom onomastics has described (Janzén 1947) and shows that it is visible in the runestones' relations, not only in single examples. That it is no weaker on the Christian stones suggests that the Nordic naming custom was alive also in the families that marked their Christian faith.

Weaknesses, in order of weight.

1. The district's stock of names. Neighbours share names, and part of the excess may be local fashion. For parent–child the finding remains within the hundred; for siblings it shrinks (5.4).
2. The division into elements is crude. It rests on the material itself and gives some junk elements. Without short and junk elements the finding remains, but hypocoristic forms like Holmi count as single-element names and may hide inheritance.
3. The relations are read by rules. The errors are about one in twenty in the check.
4. The children are adults. The sponsor is usually an adult, so the naming lies 30–50 years before the stone's style group. The comparison over time thus concerns name-giving around the year 1000, not the stones themselves.

What would overturn the finding. If the excess disappeared with a shuffle within the parish, once the material is large enough for it, or if a hand coding of the elements gave the same share for parent–child as for random pairs.

On the runestones parent and child share a name element in 5.3 percent of the pairs, against 2.8 percent with random pairing, and it withstands every sensitivity analysis. Children inherited elements of their parents' names, but only in a minority of families, and the custom did not weaken during the runestone period.