A random discovery as I was studying on Sindarin/Noldorin phonology.
David Salo marked S. taus "thatch" as an irregular development from its primitive form *tupsē (GS:287). The primitive p would first become a spirant and then a vowel, producing ui in normal Sindarin development (GS:§4.81 §4.82, §4.91; pp. 44-46).
However, there are a couple of examples in the Etymologies where primitive -se become -sa in Old Noldorin: *khyelesē > khelesa > ON. kheleha (Etym:KHYEL) and *peltakse > ON. pelthaksa (Etym:PEL). Ordinarily such a change would have no effect, since the final vowel would be lost either way. However, in the cause of *tupsē > ON. tuɸsa, the final vowel would cause a-affection before vanishing, producing toɸsa.
In this case, when the spirant became a vowel, it would produce tousa under regular development (GS:§4.82, §4.90; pp. 44-46). The late diphthong ou would develop into au as usual, making taus a regular development from tupsē. All we need is the extra phonetic rule of -se > -sa in Old Noldorin/Sindarin.
This change did not apply to primitive -sse, which remained: ON. (s)pharasse (Etym:SPAR) and also *tyulussē > N. tulus (Etym:TYUL) where no a-affection occurred.
Paul Strack Mar 29, 2015 (16:54)
David Salo marked S. taus "thatch" as an irregular development from its primitive form *tupsē (GS:287). The primitive p would first become a spirant and then a vowel, producing ui in normal Sindarin development (GS:§4.81 §4.82, §4.91; pp. 44-46).
However, there are a couple of examples in the Etymologies where primitive -se become -sa in Old Noldorin: *khyelesē > khelesa > ON. kheleha (Etym:KHYEL) and *peltakse > ON. pelthaksa (Etym:PEL). Ordinarily such a change would have no effect, since the final vowel would be lost either way. However, in the cause of *tupsē > ON. tuɸsa, the final vowel would cause a-affection before vanishing, producing toɸsa.
In this case, when the spirant became a vowel, it would produce tousa under regular development (GS:§4.82, §4.90; pp. 44-46). The late diphthong ou would develop into au as usual, making taus a regular development from tupsē. All we need is the extra phonetic rule of -se > -sa in Old Noldorin/Sindarin.
This change did not apply to primitive -sse, which remained: ON. (s)pharasse (Etym:SPAR) and also *tyulussē > N. tulus (Etym:TYUL) where no a-affection occurred.