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

===== Three big biases: =====
# You know Germanic + Romance much better → You spot foot/ped, heart/cor, two/duo instantly. → You don’t yet have Irish/Welsh in your bones to the same degree; a lot of cognates are masked by heavy Celtic sound shifts and broad/slender effects.
# Celtic data is historically sparse - For Gaulish, Celtiberian etc. we have fragments, not a full lexicon. - If we had them like we have Latin + Gothic, many more Celtic–Italic–Germanic patterns would be obvious.
# Closeness ≠ number of “obvious lookalikes” - Closeness is about shared innovations: things that are too strange to be coincidence and too specific to be simple retention. - Those are mostly in inflectional paradigms and sound laws, not in the 500 most transparent dictionary entries.