I’ve been thinking about your initial comment, namely, that the language contact tends to level out features and grammatical complexity tends to arise during lengthy periods of linguistic isolation. (Thanks, btw, for making that point — it has led me to clarify some of my assumptions.)

I have two observations. First, there is no evidence that the Nanti evidential system arose due to language contact — quite the opposite in fact. It appears to be a recent and independent innovation. Although I did not make this point clear in my post (but check out the handout I link to in the post!), I believe that Nanti evidentiality mostly resulted from the grammaticalization of extant lexical resources.

One question that interests me is what drove this relatively sudden grammaticalization. If we take the common position that grammaticalization results from token frequency, then the question becomes: Why did Nantis start to use evidential resources (e.g. verbs of perception) with greater frequency some 150-100 years ago? If, as people like Brian Joseph have suggested, evidentials are a resource for avoiding social conflict, then I am led to speculate about why social conflict would suddenly have become so salient to Nanti speakers at about that time. So you can see why I found Gnerre’s comments so interesting (quite apart from the actual facts of the Shuar/Achuar case — but yes, one solid data point alone is shaky, as you say). So I am actually imagining social processes as possibly driving grammaticalization as a distal cause. Who knows, of course, whether this is correct or not, but I just wanted to clarify that I was not thinking about language contact as the motive force behind the development of Nanti evidentiality.

Second, it seems to me that notion of social upheaval playing a role in grammatical change is actually quite in line with the notion of punctuated equilibrium, both as articulated by Dixon, and as originally conceived in the context of biological evolutionary theory. I realize you did not explicitly rule this possibility out in your original comment, but I would be curious if you have much sympathy for it, or not.

Not according to my analysis. Of course there are strategies for marking source of information, but not a grammaticalised paradigm. I haven’t yet looked at Shuar and Achuar in much detail, so I would be interested to see Gnerre’s analysis of evidentiality marking there. His 1999 grammar of Shuar doesn’t go into much detail regarding verbal morphology.

Thanks for the input. When Gnerre made his comment, I immediately thought of you, and wondered what you would make of it. I should write to Gnerre and ask him exactly what he had in mind. As you might know, he works on Shuar/Achuar, so he is probably thinking of what is going on in those languages/varieties. So, I take it that there is no grammaticalized evidentiality in Aguaruna?