Someone can ask you for something they may or may not need. You would give it to them if the need were real, and not otherwise. They would rather have it either way. So what could possibly stop them asking every time, and what would asking then tell you?
Elsewhere on this site, Wordless shows meaning appearing out of nothing. It works because sender and receiver want exactly the same thing: for the message to get through. Remove that and the whole account stops working.
Here the signaller always benefits from the resource, and more so when genuinely in need. If asking were free, everyone would ask always, hearing the request would tell the donor nothing, and no donor would act on it. A costless signal in the presence of conflicting interests is not a weak signal. It is not a signal.
Zahavi's handicap principle proposed the escape: make asking expensive. Maynard Smith (1991) wrote this game to test it, and named it for the poet who, dying of his wounds, is said to have passed his water to a soldier with the words “thy necessity is yet greater than mine”.
This is the objection that has to be answered before any evolutionary account of meaning can be believed, and answering it turns out to be harder than the standard story admits.
Skyrms's account of meaning, which Wordless models, works because the sender and receiver want the same outcome. Nothing there is tempted to lie, because lying gains nothing. That is a real result, and it is also a special case. Most communication that matters happens between parties whose interests only partly overlap.
Once they diverge, cheap signals unravel. Anyone can send them, so everyone does, so they carry no information, so nobody attends to them. That reasoning is short and it is why the handicap principle was taken so seriously.
Make the signal expensive. If it costs more than the resource is worth to someone who is fine, only the genuinely needy will pay, and the cost itself does the work that honesty cannot. The peacock's tail, the stotting antelope, the expensive engagement ring: all of them, on this account, are honest precisely because they are wasteful.
Nobody in this model is truthful. The signal is reliable because lying is not worth what it costs.
The first is a ceiling. A signal can be too expensive: past the point where someone in real need still gains by sending it, nobody sends anything and communication dies from the other end. Honesty lives in a window, not on a slope, and “costlier is more honest” is simply wrong as a summary.
The second is worse for the principle. Huttegger and Zollman showed that stable partly-honest states exist at signalling costs far below what the principle requires, and that states of no communication are stable too and often more likely to be reached than the costly honest equilibrium. Both of those show up in the runs above. What the model delivers is not “cost makes honesty” but something messier: cost makes honesty possible within a range, and even inside it a population may end up somewhere else entirely.
There are two conditions here, only one signal, and no possibility of a signaller who is sometimes truthful and sometimes not by choice. Relatedness stands in for every kind of shared interest, which flattens kinship, friendship, and repeated dealing into one number. And nothing here can lie strategically, because nothing here can represent anything at all.