It is known that the celtic languages have mutations, for instance:
Welsh:
*transcription depicts North Welsh dialects
• normal form: Cymru [ˈkəmrɨ̞] (Wales);
• soft mutation: Gymru [ˈɡəmrɨ̞] (ex.: Gwelais i Gymru "I saw Wales");
• nasal mutation Nghymru [ˈŋ̊əmrɨ̞] (ex.: Yng Nghmru "in Wales");
• and aspirated mutation: Chymru [ˈχəmrɨ̞] (ex.: Brasil a Chymru "Brazil and Wales").
The system development is fascinating, however, something intrigues me: why do only feminine nouns (at least in welsh, except the ones beginning in <ll> or <rh>) undergo soft mutation after the article? What has triggered it?
(Ex.: cath (cat) [kaːθ] and y gath [ɪˈɡaːθ](the cat)
Because as far as I am concerned, it is more usual to indo-european languages to show gender at the end of words. I thought this gender classification could be a later development but it wouldn't explain why some words starting with <ll> and <rh> were left as feminine.
Edit: it just occured to me that it could be due to the "original" feminine article (I mean its etymon).