I am obliged to ianr for that reference, even allowing that Mikhail, clearly a linguistics scholar, is answering a slightly different question. As he says, given that Sardinian has evolved away from Latin less than any other Romance language, as well as preserving many features of Latin, the shepherds in question were doubtless from Sardinia. The experiment's rationale extends from the fact that any human infant acquires an articulatory phonetics (a performative competence) specific to the language she or he is learning, with its defining phoneme inventory. This has the broader effect that she or he will then prefer or find easier certain sounds, combinations and longer patterns, while others become difficult or unintuitive. Certain sound distinctions will be inaudible, others insisted upon. Thus, although incomprehensible to these readers, the Latin texts would have been sufficiently similar to Sardinian that, employing their particular articulatory competence, the shepherds would produce a spoken Latin which one might at least hypothesise was much closer to the chatter of Ancient Rome than the speech of the classics master at Eton. It is as if, rather than written text preserved in a cache of old documents, this historically isolated speech-community has (putatively) preserved (quite unintentionally, and with a little drift) the 'how to' for an ancient spoken language. Research interest extends from pronunciation, nasalisation of vowels and so on, to elisions and slurring, and the characteristic rhythms and pauses given to longer phrases, sentences and extended units of language. I am sorry now to turn from thanking ianr to criticising him. I broadly understand the philosophical kite he wants to fly. The specific answer to how can we know is found in all that unending forensic interdisciplinary effort precisely to try to know: reverse engineering linguistic evolution through documents, synthetic linguistic modeling and so on, analysing transliterated Latin in other language texts, and clues in written Latin, particularly poetic meter. The meta-level response to your epistomological whimsifying is that you, not I, introduced the possibility of 'knowing' how Latin was once spoken. No-one presumes that 'know' in this context means the same as knowing something straightforwardly in one's everyday life. That being so, what is the meaning of your second question ? If one knows something approximately or vaguely in everyday life, one normally has the means at one's disposal to obtain better, clearer or additional information. In this context, one could indeed use evolutionary linguistic modeling and digital technology to tweak some of these verbal recordings and tentatively - always tentatively - get closer to what one proposes was ancient spoken Latin. I support your deconstructions, your postmodern scepticism, as a means of interrogating our 'reality' and the assumptions of any intellectual practice. This is enriching. But when know-nothing-ism wishes to disrupt and replace that work, then our ways must divide, you into sterile masturbatory solipsism, myself to continue, naively no doubt, to be a very tiny part of that ongoing attempt to understand everything, not least our human past. As I say, I am sorry to make this divided response to ianr. I thank him again for pointing out that reference to me. And PeckhamRose (who is, I put it to you, your mum) thinks you're a credit. I have been far too feeble about pursing this matter online and will take this it again. The point of this thread may be at an end, Ariadne . . . Lee Scoresby.