Cainntear wrote:Wikipedia cites Miguel Vidal as a source for the claim that
an original -t- would have, by normal sound laws in the local Romance languages, developed into -d- and I've not been able to read the original source, but my problem with that is that a Latin intervocalic T would naturally undergo diachronic lenition to a D (sonorisation), but the hypothesis that it started as a variant of castellum/castella... well, that would mean it was a non-intervocalic T anyway, and if the process of lenition by sonorisation was already effectively complete by the time the S was dropped, well, there'd be no need to lenite it away.
The source says:
4. Castellani
An old theory, defended by Ocampo, Zurita, Andreu Bosc, Balari and Rubio García, connects català
with castellanu through a reduced form *castl(l)án-. The regular development can be seen in cat.
Castella and castellà, which obviously refer to another part of Hispania. In Catalan toponymy there
certainly are cases derived from a reduced form castl(l)an-, but the possible developments are either
catllà /kaʎʎá/ or carlà. A development *castlan- > català is quite impossible. Rubio García argues
that the development originally occurred in the South of France, to account for the loss of s before
the consonant cluster. It is true that the modern Occitan dialects have lost s in this position at the
northern fringe of the dialectal area, but that surely happened in modern times. Around 1100, when
the word Cathalonia first appears, the s had not even been lost in the langue d’oïl.
Cainntear wrote:For example the month setembre dropped the P but didn't voice the T.
Spanish 'siete' from Latin 'septem'. Those consonant clusters were regularly simplified (unless they were learned directly from Latin, I suppose).
Cainntear wrote:Even without the non-intervocalic thing, Ts can be retained: compare Catalan "botiga" with Spanish "bodega"; and then there's more recent borrowings like "patata" with its two intervocalic Ts.
That's why you have to have a specialized expertise, to know what kind of phonological developments were there at which time periods (and when did the word appear/was borrowed from a different language, which allows it to circumvent some of the phonetic changes) to assess the possibility of this or that transformation. If some kind of authority (or someone pretending to be an authority) says something along the lines of "this development is quite impossible and should have given this reflex", then I find myself unable to determine the validity of that by myself.
'Patata' was borrowed from Spanish, so why should it undergo the voicing? In Spanish it was itself borrowed from Taíno, which is a New World (Caribbean) language, the contact with which could not have happened until the 15-16th century. There are plenty of late learned borrowings from Latin which preserve all the intervocalic voiceless stops.
Don't know about the 'botiga', how it developed in Catalan. In Spanish there is also 'botica' (with both 't' and 'c' unvoiced), which is a borrowing from the French 'boutique', and a doublet of 'bodega'. The Catalan 'botiga' seems to be stuck in the middle of these two possibilities. The original word was Latin 'apotheca' (from Greek, of course), who knows about that "th"?