Gemination in Sanskrit and Pali

Question

The Thai vocabulary consists of a lot of Pali and Sanskrit words. Their spelling is preserved, but not their pronunciation (so that for example, originally dental and retroflex stops except those both voiced and unaspirated become [tʰ] in Thai). However, in a school-level Thai language class, spelling is sometimes taught based on the lost distinction so that students understand consonant combination and redundancy.

I remember being taught that, in a Pali or Sanskrit word, a succession of two stops with the same articulation is either

• geminated unaspirated stops (eg [pp] and [tt]), or
• an unaspirated stop followed by an aspirated counterpart (eg [ppʰ] and [ttʰ]) like in 'buddha' (enlightened) and 'yuddha' (war),

but never

• geminated aspirated stops (eg [pʰpʰ] and [tʰtʰ]), nor
• an aspirated stop followed by an unaspirated counterpart (eg [pʰp] and [tʰt])

This has come up in a discussion between my friend and I, both of whom are native Thai and remember being taught the same thing. And then we question it.

How true is that in Pali and Sanskrit? Are there counterexamples? Could you please point us to an authoritative source regarding this?

Answer 1

This is most likely true. Voicing and aspiration does not contrast before an obstruent (they are written with the same voicing as the following C, and are written as unaspirated). It's most likely that Buddha etc did not have two releases, but we can't tell about breathy vocal fold abduction during the closure of the first consonant, since that requires physiological investigation. Sanskrit is described in Whitney's Sanskrit grammar which is considered authoritative. Actually authoratative information would come from the Shikshas, especially the Pratishakhyas (composed when the language was still spoken), but it isn't trivial to interpret ancient phonetic descriptions. However, the rules of sandhi promulgated by the ancient grammarians are consistent with the standard spelling (e.g. buddha), evaluated as phonemic statements. For Pali, I think the standard work is Charles Duroiselle's A Practical Grammar of the Pāli Language: there does not appear to be a comparable ancient indigenous source of phonetic description.

There are a few counterexamples in the texts, and Whitney mentions Rg Vedic akhkhalī, jajhjhatī. The generalization that you offer is correct as far as it goes but for Sanskrit insufficient, since one can also not find aspirates before any obstruent, so *akhṣa, *doghdhum.