@lars But for languages on different branches of the same [super-]family, how can we be sure that there’s no common ancestor upstream? Like for example: Proto-Germanic and Middle Persian, or Proto-Germanic and Latin?

Small note, regarding English/German cognates with other Indo-European languages:

Unless you’re dealing with a Latinism or Hellenism, distrust any potential cognate starting with the same stop or fricative. They almost never match, because of a bunch of really old sound changes that Proto-Germanic went through:

• PIE *p t ḱ k kʷ → PGerm *f θ h h hʷ
• PIE *b d ǵ g gʷ → PGerm *p t k k kʷ
• PIE *bʰ dʰ ǵʰ gʰ gʷʰ → PGerm *b d g g gʷ

Those changes are collectively known as Grimm’s Law. (After Jacob Grimm - yup, the one from the fairy tales.)

So for example. If you find a Spanish word starting with /d/, and you want to find English cognates, don’t look for English words also starting with /d/, but with /t/: dos/two, diez/ten, diente/tooth. Spanish “dia” for example does have an English cognate with /t/… and not much else - it’s the /t/ in “lent” (from Proto-West-Germanic *langatīn spring - see that *tīn?).

If you try the opposite, you’ll probably find Romance cognates with /f/ for English /d/. With Spanish then changing /f/→/h/→Ø, as in dough vs. heñir (to knead), both from PIE *dʰeyǵʰ- (to mould, form, build, knead).

The reason that island is spelt with an s is because English scholars mistakenly assumed that it was from the French isle (from Latin insula) and spelt it isle-land, which became island. It’s actually from Old English iegland.

Wait, Italian “ciao” is derived from (medieval) Latin “sclavus”/slave?