Mfcttrf

In an old comment on another question, jlawler mentions in passing:

Much the same can be said about ejective consonants -- other languages can pick them up, but nobody knows where they come from. Clicks and ejectives are common in the phonologies of Sprachbunds, and nobody knows where they come from, either.

This took me by surprise, since I would have expected ejectives to arise naturally from e.g. clusters of stops plus [ʔ], or pharyngealized consonants becoming glottalized, and so on.

However, I certainly can't think of any instance of ejectives arising in this way in a language that didn't previously have them.

So—is this comment correct? Are there really no known instances of ejectives developing naturally from non-ejective consonants, rather than spreading in from an adjacent language?

It is almost true, in the sense that there are nearly no cases of ejectives unambiguously developing and clearly without external influence. There are two good candidates, though: Yapese and Waimoa. Ethiopian Semitic development of ejectives can be attributed to contact with Cushitic, and Nguni ejectives to contact with Khoisan. However, it is also possible that proto-Semitic originally had ejectives and pharyngealization is a later development. A further however is that South Arabian languages also have ejectives which would seem to be inconsistent with a Cushitic contact hypothesis, but it is also conjectured that Cushitic languages were originally spoken in South Arabia. In other words, the possibility of contact can't be clearly ruled out.

There are three and two fractions candidates for sua sponte ejectives. The first fractional candidate is Korean: but it is generally agreed that the glottalized consonants of Korean are not ejective, they just have some amount of laryngealization. The second fractional candidate in Tonkawa, which either has [tʔ] clusters or ejectives. Phonologically they behave like Cʔ clusters, and Hoijer's description of their pronunciation doesn't make them sound like Salishan ejectives (but Salishan ejectives are perhaps too architypical). There are no prospects for phonetically investigating Tonkawa since there are no sound recordings and the language is extinct (but also no related languages so even if there are ejectives, they did not demonstrably develop from something else). These languages represent a further complicating factor, where "glottalized" non-voiced consonants may not be phonetically ejective.

Yapese, Waimoa and Itelmen are reported to have ejectives. The problem with Itelmen is that it's not clear what the historical predecessor was, so this could just be a retention (Chukotkan languages don't have ejectives – but they might not be related, or they might have lost original ejectives). The other two cases are Yapese and Waimoa, both Austronesian languages, where there is no suggestion that proto-Austronesian had ejectives – and these languages are not in contact with any language with ejectives. Blust ("More on the origins of glottalic consonants") doesn't have much to say about their source in Yapese, except a few cases where they may have developed from C+ʔ sequences. Maddieson reports phonetic results for Yapese ejectives, which indicate that the putative ejective stops (but not the fricatives) are indeed ejective, not just C+ʔ sequences like in Tonkawa.