At the risk of continuing a topic where there is clearly passionate disagreement:

I cannot resolve this percentages disagreement in my own head: I lack the data and I'm not clear about how one could conclude that a broadsheet/print origin was a fact. Just because a printed version is the earliest known version it does not mean there were no oral versions before this. I guess this is maybe where the Wittgenstein bit comes in.

Not quoting pages numbers this time, but I think Roud does acknowledge that there may be oral origins for some of these, but says if so how we can we know? The quality of language/aesthetic judgements have been suggested as one way of deciding, but this is a very subjective area.

But unless I am wrong, Roud's book is intended to apply only to a 'corpus' collected up until the mid 20th century, and collected mainly by the people whose stories Roud tell us, and the percentages apply to this corpus, as represented mainly by the songs within that category to which Roud has given numbers. Have I got this right?