some thoughts of mine:
We have to be cautious about presence or absence of cognates words in languages
1- old words can have disappeared (and recently enough sometimes)
2- the litteral meaning of old words can evolve, even if giving birth to new meanings close enough or still related to old meanings (see the exchanges between meanings like «horse», «stallion», «mare», «foal» or «colt», «filly» ...)
3- on another side, without shift of meaning, by instance, in a population where horses are common and well used for different purposes, a lot of names can exist for them
I agree with that, such evolutions are indeed a reality and I don't deny it. The problem is rhetorical: we have a PIE etymon A and its reflexes in the attested IE languages (let's call them A1,A2,A3 etc). We have thus three possible situations:
1- The meaning of the reflexe A1 is the same as the meaning of the etymon A :
A = A1 (and possibly A2, A3, etc). Ex: ek
wos = equus + hippos (horse = horse) or *g
he → go, gå (to quit → to go)
2- The meaning of the reflexe A1 is not the same as the meaning of the etymon A BUT you can observe similar distorsions in other IE languages' reflexes (A2,A3, etc).
A → A1,A2,... . Ex: PIE *b
hreus (to blow up) → Gaulish brunnio (breast) ↔ PGmc *brustiz (eng. "breast", dan. "bryst" etc.)
3- The meaning of the reflexe A1 is not the same as the meaning of the etymon A AND you CANNOT observe any similar distorsions outside the language considered :
A → A1. Ex: *d
hereg (to hold) → Danish dreng (=boy).
To put it simply : In 1 the PIE etymon is clearly reliable, in 2 its likelihood is acceptable although not certain (it can be a borrowing, a wandering word, or a non-IE etymon) and in 3 is speculatory. Unfortunately, most PIE etymons belongs to the last category.
You have two schools, as in the reading of the Bible : the "
maximalists" who think that, since all the attestable historical events of the Bible are true (eg the exile to Babylon) then all the non-attested historical events should be true as well, and the "
minimalists" who think that only the attested events are true.
In historical linguistics, you have the same dichotomy : people who believe that since processes of semantic distorsions are "attested" and true, non-attested semantic distorsions should be true as well. And people who believe only in what is attested. From this perspective I am clearly a minimalist : I agree with 1, sometimes with 2, never with 3.
Your example is at the very edge of 2 and 3. As long as we are talking about horses, I agree with the connection horse/stallion, not with horse/mare. As for myself, and for most of us I suspect, I couldn't make the difference between a horse and a mare and I wouldn't care riding either of them (if I could ride a horse). But for people from the Neolithic, the probability that they could have mixed the two seems very unlikely.
You have such alleged evolutions in French with Vulgar Latin, as f.ex. with *pulla (offspring - of an animal) → poule (hen) , poulain (colt), poutre (beam). Maybe it is true, but to me it is pure guesswork.
Moreover, I don't believe at least in two semantic distorsion processses which are :
- Echoism
- Taboo words
If you discard the situation3, echoism and taboo words, the remaining lexical stock is rather small. I agree with you and Taranis about the existence of *ek
wos reflexes in Celtic. But "cabalos" or "marc'h" are obviously not IE, which shows that the PIE horse culture is not specific.
We should have a thread devoted to this horse question, since it is in fact so critical within IE studies.