What it's really about is the presence or absence of regular sound
correspondences. So let's say you have a sound change that only remains
active for a limited time and doesn't have time to affect all of the
vocabulary. After it has done its job you are left with a certain amount of
irregularity in the language evolution. You will also often have loaning
between dialects or closely related languages which makes for words that
can gather an idiosyncratic set of sound changes while bypassing others.
You might also want to consider more or less irregular contractions of
commonly used words.

This all easily creates words which have similar forms in the daughter
languages but because of the irregularity of the changes it might be
difficult to prove the cognancy. You can stir the pot even further by
having adjacent dialects develop similar affective vocabulary under the
influence of each other. Here you get a similar problem of cognancy with
the resulting words being now related but not cognate.


2014-08-08 1:18 GMT+03:00 Anthony Miles <[log in to unmask]>:

> No natlangs are 100% cognate with another language. But as a Conlanger, I
> can design a language and be certain of everything which is cognate. This
> includes items which, if it were a natural language, a linguist might not
> be sure if it were cognate or not. My question is: how different most a
> form become before it no longer comes as a cognate form? This is clearly a
> question which is a byproduct of our art, but I was wondering if anybody
> had given any thought to this issue.