How is son reconstructed in PIE, sou̯nus or seu̯nus?
Starostin gives contradictory accounts: in Indo-European etymology page he gives the first variant, but on a page for Eurasiatic etymology he gives the second one. Possibly the vowel could be alternated in this root depending on meaning (to give birth/son), similarly to alternation in ret-/rot-.
Morphologically, we have CeRC here (R stands for the resonant).
It’s an athematic noun, which means that endings were added directly to the stem (sunu-) without an intervening thematic vowel. Obviously, it is also an u-stem.
All athematic nouns have strong (e.g. nom. and acc. sg., nom.pl.) and weak cases (the rest).
If you subscribe to the accent and ablaut paradigms, this particular noun would be proterokinetic (or proterodynamic), which means that in the strong cases the accent was on the full-grade root (the e-grade), whereas in the weak cases both the accent and full grade shifted onto the suffix, so the u-suffix becomes –eu- (or –ou-, the o-grade) in the weak cases. So, at this stage, we reconstruct NOM.SG. as *seunus - the standard citation form is the e-grade.
However, in the PIE daughter languages the root didn’t ablaut already, cf. Sanskrit sunus, Lith. sunus etc. (sorry, no diacritics).
Now about the root vowel. Usually the long u is not reconstructed as a vowel phoneme in the PIE. Since in Sanskrit and Lithuanian we have a long root vowel – and “y” in Slavic, we need something to make our short u long. If you subscribe to the laryngeal theory, obviously you propose a laryngeal H (capital H means unspecified); some researchers propose h3 here. So, now, at this stage, we reconstruct NOM.SG. as *seuHnus (or *seuh3nus). For the weak cases, we could reconstruct the zero-grade stem *suHn-eu or rather *suHn-ou (other variants: *suh3neu, *suh3nou). Here's a partial paradigm for this noun (singular) proposed by Leonard Herzenberg (1934-2012) (sometimes his last name is also transliterated as Gercenberg):
IE *séu̯h3-nu-s ‘son’
NOM. *séu̯h3-nu-s
ACC. *séu̯h3-nu-m
DAT. *su̯h3-nóu̯
GEN. *su̯h3-nóu̯-s
(Herzenberg 2010: 69)
Re: the root vowel in the strong cases
Once again, in the accent and ablaut paradigm theory (a la Leiden, not Erlangen??), if a noun is proterokinetic, in the strong cases the root is in the full grade and accented, whereas in the weak cases the suffix gets both full grade and accent (e.g. Fortson 2010, §6.22). In other words,
for the strong cases we would have R(é) S(∅) E(∅), but for the weak cases we would have R(∅) S(é) E(∅) (Weiss 2020, p. 277),
where R stands for root, S suffix, E ending, é full accented grade, ∅ zero grade.
Thus, we have
nom.sg. CéC-R (s), acc.sg. CéC-R (m), gen.sg. CC-é/óR-s etc. (Beekes 2011, 2nd ed., revised and corrected by M. de Vaan, p. 190).
There are at least two ways to reconstruct PIE 'son' under this theory.
They are explained in detail in e.g. Meier-Brügger 2010 (9th ed.), pp. 336-338 (F 314, 1 and 4) and pp. 341-342 (F 317, 1-4)
"Vgl. ferner aus dem Paradigma von ved. ’Sohn’ Nom.Sg. stark sūn-ú-s (das Suffix ist schwundstufig, der Akzent lag ursprünglich auf der Wurzel, ist aber sekundär auf das Suffix verlegt worden, s.u. Abs. 4), Gen.Sg. schwach sūn-ó-s < *sūn-áu̯-s (das Suffix ist vollstufig)." (emphasis mine - Alex B., p. 336)
His reasoning is as that even though on the surface level "Das Paradigma von ved. sūnú- zeigt durchgegend stabilen Akzent auf dem Suffix", [d]ieser Zustand kann aber nicht alt sein. Der Akzent ist auf dem vollstufigen Suffix des schwachen Stammes des Erwartete, nicht aber auf dem schwundstufigen Suffix des starken Stammes" (p. 342).
Meier-Brügger also writes that the problem of the root vowel in the strong cases in PIE 'son' cannot be settled, that is why he gives the following reconstruction for the strong cases, " uridg. *?-nu-s, whereas for the weak cases he gives "uridg. *suH-néu̯-s", cf.
"Während der Ablautwechsel beim Suffix ved. einzelsprachlich belegt ist, fehlen solche Hinweise auf den Ablautwechsel bei der Wurzelsilbe sū-. Die Rekonstruktion muß dewegen in diesem Bereich unsicher bleiben. Alternativ zu vollstufigen *séu̯H- kann auch betontes schwundstufiges *súH- nicht ausegeschlossen werden." (p. 342)
Cf. the following remark from Fortson 2010 (p. 125, §6.42):
“Two classes of athematic nouns that behaved in parallel were those whose suffix was *-i- or *-u- in the zero-grade and *-ei- *-eu- in the full grade. The most common type continues an earlier proterokinetic paradigm, with zero-grade *-i-s *-i-m and *-u-s *-u-m in the strong cases, and full grade in the weak cases (genit. *-ei-s *-eu-s). By the time of the daughter languages, however, these nouns do not ablaut in their root syllable.”
If I understand Fortson 2010 correctly, he argues that in such cases the zero-grade (for the root in the strong cases) has been generalized (analogically restored) from the weak case forms with the ablauting suffix.
Of course, you could rather agree with Jesse Lundquist (Lundquist 2018) instead, who wrote that
"Because the hypothesized changes are situated deep in prehistory (i.e. "prior to PIE as accessible by the comparative method"), their plausibility is difficult to evaluate, either within individual classes or collectively, at the systemic level (p. 2135), cf. NIL *suH-
If you don’t subscribe to the laryngeal theory, you could use a schwa instead of the laryngeal or simply reconstruct *suənu-, without any ablaut.
Pokorny also gives the seuə- account:
seuə-1 To give birth. Suffixed zero-grade form in derivative noun *su(ə)-nu-, son. son, from Old English sunu, son, from Germanic *sunuz.
