The origins of morphological irregularity

Consider a language with a verb ak ‘to be’. This language inflects its verbs for person like so:

• 1st person: -i
• 2nd person: -to
• 3rd person: -a

This gives us the paradigm in Stage 1 of the language:

• ak-i ‘I am’
• ak-to ‘you are’
• ak-a ‘he/she/it is’

But imagine this language undergoes palatalization, so Stage 2 looks like:

• atʃ-i ‘I am’ (k > tʃ / _i)
• ak-to ‘you are’
• ak-a ‘he/she/it is’

Then it undergoes consonant assimilation, yielding Stage 3:

• atʃ-i ‘I am’
• at-to ‘you are’ (k > t / _t)
• ak-a ‘he/she/it is’

Then unstressed vowels delete, with the following result in Stage 4:

• atʃ ‘I am’
• at ‘you are’ ([t] also degeminates)
• ak ‘he/she/it is

So by the operation of a perfectly regular process of sound change, we produce a very irregular verb! See [[ Sturtevant’s paradox ]].