Re: [tied] Artificial Languages (reminiscent of IE)

I originally didn't want to post this b/c I thought it was off-topic. But
apparently, that isn't a concern with this discussion group. (Yep, I'm
talking about the Egyptian mythology, etc.)

I have dabbled in artificial language construction for an imaginary world
(yeah, I know that makes me even geekier than amateur linguists.) I prefer
to create fictitious languages that mirror the 'realistic' qualities of
natural languages, for example, irregular forms, numerous/unusual forms, and
diachronistic qualities, such as phonological, lexical, and grammar changes.
Anyway, I'm definitely NOT a professional linguist and my smattering of
exposure has waned thin in the past decade.

In the interest of creating a 'realistic' artificial language or as much as
I may with my current knowledge and time, I would like to know how much I
should comply with known 'rules.' In other words, can I just pick and
choose sound changes without too much correlation as long as there is a
general verisimilitude of consistency or would that be blatantly
unrealistic? It seems that this would be acceptable since what I garner
from the discussion suggests that there is much debate among professional
lingustics about what rules govern changes and characteristics of natural
languages. There are many words and and reconstructed forms that members of
this group have posited--consensus has not been unanimous at all--even with
regard to the 'rules' for the differences and apparent inconsistencies of
natural languages.

To put it in more concrete terms, let me give an example. The parent
language has the word 'smé' meaning 'three.' The daughter language has the
word 'smía,' also meaning, of course, 'three.' 'Ré' in the parent language
means 'human, man.' Am I obligated to make the daughter language's cognate
form be 'ría' or can I make it 'ru' without any other basis or forms to
provide support an irregularity? I realize this is an
oversimplification--many rules can operate, but I hope it serves to
illustrate my point. In other words, do such 'glitches' and inconsistencies
occur in natural languages? If so, how often? How much leeway do I have?
I realize no artificial language will be able to imitate fully the true
verisimilitude of a natural language--but I wish to attain a convincing
fiction--ie, no obvious flags that tell a linguistic, 'This is horribly
bogus.'

Thanks,
Kraig