Crane signalling

Why does a crane crew let only one person give signals, yet allow anyone at all to call a stop?

On a lift, the rule is strict: one appointed signaller, and the operator takes instructions from nobody else. A second person waving is not a helpful contribution — it is a hazard, and the operator's correct response is to stop.

And then the same rule book says that anyone on site may call an emergency stop, and the operator must obey it immediately, without checking who it was or why.

Those two rules look like they belong to different philosophies. One centralises authority absolutely; the other distributes it to everyone including the least experienced person present. Yet they sit in the same procedure, and they are not a compromise between two camps. They are two answers to two genuinely different problems, and the asymmetry between them is the whole point.

Start by separating what the two rules are protecting against.

The single-signaller rule protects against conflicting instruction. Consider what a signal actually is: it commits a large mass to a motion, and the operator often cannot see the load, the landing area, or the people near it. They are acting on someone else's eyes. If two people signal and they disagree, the operator has no way to adjudicate — both look authoritative, both may be responding to something real, and the operator lacks the view that would let them choose. Worse, the operator may blend the instructions, or hesitate, and hesitation with a swinging load is itself dangerous.

So for positive action — move, slew, hoist, lower — the requirement is a single source of truth. Not because the signaller is the most senior person, but because the instruction stream must be unambiguous. This is why the appointed signaller is identified visibly before the lift, why they alone are in radio contact, and why the operator stops if they lose sight of them: a signal from an unknown source is not a weaker signal, it is a corrupt one.

The stop rule protects against something else entirely: nobody having the right view. The signaller has one vantage point. They cannot see everything. A person walking under the load, a sling slipping, an outrigger sinking, an obstruction the crane is about to swing into — these may be visible to a labourer at the other end of the site and invisible to the signaller and the operator both.

Now weigh the costs of the two possible errors for a stop call. A stop that turns out to be unnecessary costs some minutes and some embarrassment. A hazard that is seen and not acted on because the observer was not the appointed signaller costs a great deal more, and it is precisely the error a strict hierarchy would produce — somebody sees the problem, hesitates because it is not their place to speak, and by the time it reaches the signaller it is too late.

That asymmetry decides it. Where the costs of the two errors are wildly unequal, the authority to act should sit wherever the information is. For starting motion, the danger is conflicting input, so authority is concentrated. For stopping motion, the danger is missing information, so authority is distributed.

And notice the structural elegance: a stop instruction cannot conflict with another stop instruction. Two people shouting stop produce the same outcome as one. The reason the single-source rule is needed for positive signals — that instructions can contradict — simply does not apply to a command with only one possible content. So distributing it is free.

"One signaller" is about the channel, not about rank. The appointed signaller is frequently not the most senior person present, and the lift supervisor may be someone else again. What the rule establishes is that the operator has exactly one input for motion commands. Reading it as a hierarchy leads to the wrong conclusion — that authority flows from status — and that reading is what erodes the stop rule in practice, since a worker who thinks signals belong to important people will hesitate to shout.

The stop authority is fragile in exactly the way the reasoning predicts. It works only if calling a stop is genuinely costless to the person calling it. If someone who stops a lift unnecessarily is mocked, or blamed for lost time, the rule stays in the procedure and stops functioning — because the individual's cost has been raised while the site's benefit stays diffuse. Sites that take this seriously deal with it deliberately, by thanking people for stops that turned out to be unnecessary. That the rule needs active protection is evidence that the incentive runs against it.

Radio does not remove the problem, and can worsen it. It is tempting to think that clear communications make the single-signaller rule obsolete. But a shared channel with several voices reintroduces exactly the ambiguity the rule exists to prevent, and radio also removes the operator's ability to confirm who is speaking. The rule is about the number of sources, not about the medium.

Standard signal sets are conventional and vary, which is a real hazard on mixed sites. Hand signals are codified by national standards and by industry bodies, and they are not identical everywhere; the same gesture can differ between jurisdictions and between trades. So signals must be agreed before the lift rather than assumed, and a crew assembled from different backgrounds is a specific risk. This is a case where the convention is genuinely arbitrary and the important thing is that everyone holds the same one.

The falsification test. If the two rules answer two different failure modes, then incidents should divide accordingly: multiple-signaller situations should be associated with wrong-motion incidents, while incidents involving people or obstructions outside the signaller's view should be associated with a failure or absence of stop calls. If both kinds of incident arose equally under both conditions, the rules would not be tracking distinct mechanisms and the asymmetry would be arbitrary convention rather than design.