The following table shows my reconstructions of some of the forms of the declension of the word for “sister” in Proto-Indo-European, Proto-Celtic and Proto-Germanic. I have included only the nominative, accusative, genitive and dative forms (singular and plural).

 Proto-Indo-EuropeanProto-CelticProto-Germanic
nsswesōrswesūr > sweūrswezōr > swestēr
asswesorṃswesoran > sweoranswezarų > swestrų
gsswesrósswerros > sweorosswestraz > swestres
dsswesréiswerrei > sweoreiswestrī
npswesoresswesores > sweoresswezarez > swestrez
apswesorṇsswesoras > sweorasswezarunz > swestrunz
gpswesróhxswerron > sweoronswestrǭ
dpswesṛbhósswerribos > sweoriboswezurmaz > swestrumaz

The declension table appearing on the Wiktionary page for PIE *swesōr follows the amphikinetic declension model according to which the oblique forms have the first syllable in the zero-grade (*sus– instead of *swes-). I understand that there must be a reason for reconstructing this type of declension model, but I don’t see it represented in the declensions of any of the reflexes of *swesōr. (For the genitive singular –ós instead of –és, see https://vellaunos.ca/2021/09/15/about-the-genitive-singular-ending-in-pie/.)

In the Proto-Celtic declension, we see the sequence –sr– regularly becoming –rr– in the genitive and dative cases (singular and plural). This disappearance of the medial -s- in the genitive and dative caused the medial -s- to also be dropped in the nominative and accusative, resulting in *sweor– from *swesor– in the accusative singular and in the nominative and accusative plural (and possibly *sweūr in the nominative singular). The form *sweor– was then generalized to the entire declension. This form is represented in the Gaulish inscription found at Néris-les-Bains, in which the instrumental plural suiorebe (“with the sisters”) appears.

The Gaulish suiorebe has been used to support the idea that intervocalic -s- regularly disappeared in Celtic as it did in the Celtoid groups (Goidelic and Brittonic) but the disappearance of intervocalic -s- in this particular word is obviously exceptional. Two other words from Gaulish inscriptions that have been used to point to the possibility of intervocalic -s- regularly disappearing in Celtic are siaxsiou “I shall seek” (Châteaubleau line 6) and sioxti “possession” (?) (La Graufesenque – sioxti albanos pannas exra tuθ ccc).

The former word appears to be a reduplicated future of *sag– “seek” – *se-sag-sjō. The latter word may be from a reduplication of the PIE root *seghj– “possess, own, control” with the abstract nominal suffix –ti– – *se-sog-ti– (although it is hard to see how this interpretation can help the sentence make sense). In both of these cases, the disappearance of the second -s- is probably due exceptionally to the reduplication of a root beginning with s-.

Those who favor the idea that intervocalic -s- regularly disappeared in Celtic have to either find explanations for intervocalic -s- appearing in Gaulish (such as simplified -ns- and -ts-), or simply ignore instances of intervocalic -s- appearing in Gaulish.

The declension of Old Irish siur shows a medial -th- instead of the original Proto-Celtic medial -s- in most forms (genitive singular, genitive and dative dual, all plural forms). Intervocalic -s- regularly disappeared in the Celtoid group (Goidelic and Brittonic), so it appears that the -th- in Old Irish resulted from the addition of -t- (probably in Proto-Goidelic) by analogy with other kinship terms like Old Irish athair “father”, bráthair “brother” and máthair “mother”. The addition of this -t- in Proto-Goidelic probably happened after the disappearance of the original intervocalic -s-.

In the Proto-Germanic declension, we see the regular development of –str– from –sr– in the genitive and dative singular and in the genitive plural. The form –str– was generalized to the whole declension with the exception of the nominative singular – this shows –stēr by analogy with other kinship terms like *fadēr “father”, *brōþēr “brother” and *mōdēr “mother” (similar to what happened in Old Irish). (But I think it likely that the accusative singular and the nominative and accusative plural had –star– intead of the –str– that appears in the table above.)

[By the way, the genitive singular –es ending in *swestres did not develop from the –az ending in *swestraz but resulted from the extension of the a-stem genitive singular ending to all consonant stems. This a-stem genitive singular ending –es was reduced from an earlier form –esja which resulted from the following development: –esja < –asja < –osjo. I treated this issue in the following article: https://vellaunos.ca/2021/09/15/about-the-genitive-singular-ending-in-pie/]