The best way of looking at the prefixation of "t" to "s" is that it's the
particular form taken by lenition of "s" when the previous word ends in "n".
"S" can be lenited only when followed by a vowel or by l or n or r. So it's
only in the same cases that "t" can be prefixed.
Prefixation of "t" to "s" under lenition is not confined to the article. You
will also find it - at least dialectally - after "aon", "sean" and other words
which lenite and end in "n".
The easy way to remember when "s" can be lenited is that it can when the letter
following it can't, and vice versa. This is borne out by jtm's lists (apart
from "sh" but then it would already be lenited and we wouldn't be trying to
lenite it again):
> allowed: a e i l n o r u
> not allowed: b c d f g h m p s t
> never happen(?): b d g s
> masculine loan words only(?): f
> aspiration: h
> unvoiced stops: c p t
> (perhaps some rule preventing 2 unvoiced stops initial - I
> can't think of anything like "Ptolemy" in Gaelic, or for that
> matter, anything like "apt".)
> left over: m
> Does anyone know why not 'm'? 'tm' is about as difficult
> as 'tn' to pronounce. I remember reading somewhere that there
> are some dialect differences regarding whether the s preceding
> a slender m is broad or slender. s preceding a slender n,l,r,t
> is slender, but s preceding a slender p is broad. Since the loanwords
? that have s followed by a slender f also have the s broad (in F.P.),
? I guess this broad s before slender (f,p,m) is due to their
> labial characterisitc.
The "rule" about leniting "s" when the next letter can't be may be an aid to
memory but I'd hesitate to make a principle of it. Especially as l,n,r are
supposed to be lenitable in speech though not in print (Oh no, don't start
THAT one up again!)
I think it's more significant that the group of consonants which can follow
a lenitable "s" are all DENTAL, as is "t", so that the dental aspect remains
dominant under mutation. If you allowed "sm" to become "tm", you would be
changing a predominantly labial cluster into one which is part-labial, part
dental (OK, s is dental, but not as dental as t!!!) It's a fairly general
feature of mutations that they don't change the point of articulation
(be it labial, dental or velar) - the few exceptions to this are probably
Ciara/n O/ Duibhi/n.