Editing Openai/69186d20-2d14-8013-b9c6-c09000f173f6 (section)

===== Most of the obvious cognates you see across branches are just PIE common stock that everyone kept: =====
* ped- / foot
* tri- / three / trí
* new- / novus / nua
* gʷou- / bos / bó

That overwhelms everything.

Now suppose:
* PIE → Proto-Italo-Celtic for, say, a few centuries,
* then splits into Proto-Italic and Proto-Celtic.

In that narrow window, how much new lexical material gets created that:
# originates in Proto-Italo-Celtic (i.e. is an innovation relative to PIE), and
# survives in both Italic and Celtic, and
# is subsequently lost or replaced in Germanic and everyone else?

Answer: not necessarily very much, and a lot of it will be in semantic niches or later overwritten by contact/innovation.

So yes, you’d expect some extra shared retentions, but the denominator is enormous (thousands of PIE roots), and Italo-Celtic’s “own” layer is likely small and noisy. That’s why historical linguists lean heavily on systematic morphological and phonological innovations, where the signal-to-noise ratio is better.