the “something about this measurement system is generous” read is where I’d push, because I went back through the same tendon papers a few weeks ago and landed somewhere different on that exact point. you’re treating the size of the effect (treated animals indistinguishable from sham by day 10-14) as the thing that shifts your prior toward suspicion. I’d separate those. a large, consistent effect size is actually what should raise the prior, not lower it. the thing that keeps the human conclusion open isn’t that the effect is big, it’s that the provenance traces to basically one lab in Zagreb with thin independent replication. those are two different problems and collapsing them is the same move you’re (rightly) calling out elsewhere in your post. big effect + single source = raise the prior, keep the conclusion open. not big effect = generous ruler. on the systemic-effects-missing-from-logs piece, your third option (people aren’t tracking carefully enough) is doing more work than you’re giving it credit for. most user logs are a date and a dose, maybe a pain number. nobody’s logging gut motility or BP or whatever the gut-brain axis claim would actually predict, so “absence in logs” is close to no signal at all. the one thing that moved the needle for me on my own shoulder run was being able to drop a free-text note onto each injection entry, so the off-target stuff I’d never have thought to chart shows up when I read it back. without that you’re not measuring the absence of systemic effects, you’re measuring the absence of anyone writing them down. where I fully agree: mechanism plausible, rodent data suggestive, human outcome data absent, three different tiers. three years of reading this stuff and I’m in the same spot. my shoulder is about 80% back and I still can’t tell you it was the BPC vs time vs the PT, and that’s with me trying to isolate it. your RCT falsification criterion is the right ask. I haven’t found one either and I’ve looked. fwiw the oral-and-parenteral-both-working line is its own can of worms, mixed lit, but that’s a different thread.