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

===== The “shared predisposition” idea is still useful, but at a slightly more abstract layer: =====
* Proto-Celtic’s willingness to use sonorant fortition/gemination morphologically → gives descendants a kind of structural affordance: “sonorants are a place you can encode grammar.”
* Goidelic uses that to keep a fortis/lenis sonorant system for a while, then mostly lets it die.
* Brythonic uses it to push some of those fortis sonorants into voiceless phonemes like [ɬ], [r̥].

So:
* Concrete shared feature: PCelt ll, nn, rr, mm geminates in specific lexemes and morphophonologies.
* Diverging reflexes: Irish orthographic ll vs Welsh ll [ɬ] (and similar cases).
* Shared predisposition: the broader Celtic habit of turning “odd sonorant behaviour” into grammatical material, which both branches exploit in different ways.

You’re absolutely right that in the coll ~ cyll / cill ~ cell type pairs, we’re in the “diverging reflexes of a shared historical feature” regime, not mere convergence. That’s the correct phylogenetic reading.