The Welsh word for "cat" is cath. "His cat" is ei gath, "her cat" is ei chath, and "my cat" is fy nghath. The possessives ei "his" and ei "her" are spelled and pronounced the same. The only thing that tells them apart is what happens to the first consonant of the next word. Welsh grammar calls these changes mutations (treigladau), and it has three main ones: the soft mutation (treiglad meddal), the nasal (trwynol) and the aspirate (llaes).
Three sets of changes
Each mutation changes a fixed set of consonants:
- Soft: p→b, t→d, c→g, b→f, d→dd, m→f, ll→l, rh→r, and g is dropped altogether. Tad "father" → ei dad "his father". Gardd "garden" → yr ardd "the garden".
- Nasal: p→mh, t→nh, c→ngh, b→m, d→n, g→ng. Fy nhad "my father", fy ngwely "my bed".
- Aspirate: only p→ph, t→th, c→ch. Ei thad "her father", ei phen "her head" (from pen).
The word pen "head" goes through all of them: ei ben "his head", ei phen "her head", fy mhen "my head". When the next word starts with a vowel, "her" adds an h instead: ei enw is "his name", ei henw "her name".
What sets them off
The trigger is the word in front, or the grammar it belongs to. A feminine singular noun softens after the article y "the", so cath becomes y gath "the cat". An adjective softens after a feminine singular noun: cath fawr "a big cat", from mawr. The plural y merched mawr "the big girls" has no mutation. The preposition yn "in" takes the nasal mutation and changes shape to match: Cymru "Wales" → yng Nghymru "in Wales", ym Mangor "in Bangor", yng Nghaerdydd "in Cardiff". The aspirate follows a "and" (coffi a theisen, "coffee and cake"), â "with", tri "three" and chwe "six": tri chi "three dogs", chwe phlentyn "six children". Gender is involved here too. Tri is only used with masculine nouns, and a feminine noun like cath takes tair, which does not mutate: tair cath, but dwy gath "two cats", because "two" softens.
Where they came from
The soft mutation began as lenition, a sound change that weakened consonants between vowels: Latin securus was borrowed as Welsh segur, and superbus as syberw. In Brittonic, the ancestor of Welsh, this weakening also happened across a word boundary when the word in front ended in a vowel. Then, according to Gareth Morgan's Reading Middle Welsh, Welsh lost its final syllables around the early sixth century AD. Once the vowel that caused the change was gone, the weakened consonant was left as a marker of grammar. Morgan's example is *oina brigantinissa, which became un frenhines "one queen". Oina is the feminine form of "one"; in Brittonic most feminine nouns and adjectives ended in -a, which is why a feminine singular noun still softens the word after it. The nasal mutation comes from words that once ended in a nasal: fy goes back to Proto-Brythonic *mɨn. For the aspirate mutation, Kenneth Jackson's Language and History in Early Britain (1953) proposed doubled voiceless stops (pp, tt, cc) that became the fricatives ph, th, ch. David Greene, Patrick Sims-Williams and others have argued for a different sequence of changes, and Graham Isaac defended Jackson's route in 2004. On either account the trigger words once ended in a consonant. The conjunction a once had the form *ak, and that lost final consonant is what still changes teisen to theisen.
The other Celtic languages have related systems, built on different sounds. In Breton, "cat" is kazh, "his cat" is e gazh and "my cat" is ma c'hazh. Irish has a chat "his cat" with lenition, a cat "her cat" with no change, and ár gcat "our cat" with eclipsis, the Irish form of nasal mutation. Where Welsh marks "her" with an aspirate, Irish leaves the consonant alone.