↩ SocioRhetoric · A New Catalogue of Speech Acts
sociorhetoric · a speech act
Why do the most secure people skip the very signals the anxious can't stop sending — the billionaire in a plain hoodie, the expert who says "I just tinker"? When does not signalling become the strongest signal of all?
The economists Feltovich, Harbaugh, and To gave the puzzle its sharpest form in 2002, in a paper titled, with a straight face, Too Cool for School? Their finding is counterintuitive and, once seen, everywhere: the people who signal hardest are in the middle. The top abstain. A medium type needs the costly marker — the flashed watch, the named-dropped school, the title on the door — because without it he risks being mistaken for the bottom, and the marker is the one thing that pulls him clear. The very top abstains because some other channel already vouches for it: a reputation that precedes you, provenance, a competence too evident to state. Matching the mediums' effort would only pool the high type back in with the anxious.
The result is non-monotonic: as you climb from low to high, signalling first rises, then falls. The middle shouts; the top goes quiet; the bottom can't afford to make a sound. And because the whole logic runs on the second channel, the countersignal is legible only to those already tracking the game. To an outsider — someone who cannot read the reputation, the provenance, the room — the plain hoodie is not security. It is just a plain hoodie, and its wearer reads as low. The instrument below builds this out as a real three-type game and solves for the equilibrium, so you can watch the high type fall silent, then force it to speak again by draining the second channel dry.
the signature instrument
Three types — low < medium < high — each choose how hard to signal (a costly effort). A noisy secondary cue independently vouches for the genuinely high, with informativeness you control. The receiver sees both the signal and the cue, forms a posterior, and pays standing. The displayed values — the equilibrium effort by type, the receiver's posteriors, the non-monotonicity flag — are computed from a logit best-response solve of the game.
The signalling game
3 types · effort 0–4 · equilibrium by best-response iteration
The settings
Equilibrium signal effort, by type
What the receiver concludes
the loudest are the middle
Start at the bottom. The low type would love to look like more, but the marker is priced in a currency he is short of — the education he didn't get, the hours he can't spare, the polish he can't fake. Effort costs him the most per unit of standing it buys, so he stays near silent. Not chosen silence; forced silence. In the instrument, the low bar barely lifts off the floor.
The medium is the one with everything to prove. He is close enough to the bottom that, left unmarked, the room might file him there — and unlike the high type, he has no second channel vouching for him. The costly signal is his only lever, and he pulls it hard. This is the engine of the whole pattern: the person who most needs to separate is not the best, but the one most in danger of being confused with the worst. Watch the medium bar climb above both its neighbours.
The high type could out-signal the medium easily — effort is cheapest for him of all. He declines. The secondary cue already places him at the top, so the marginal standing a loud signal would buy is nearly nothing, while its cost, though small, is not zero. Worse, in the equilibrium the loud signal has become a medium's tell: to send it is to be read as a medium. So the high type does the one thing the anxious cannot bring themselves to do — he says nothing, and the saying-nothing, against the backdrop of the cue, is what marks him.
what to try
Load countersignalling and read the three bars: the medium shouts, the high goes quiet, the low can barely speak. The non-monotonic flag lights up — signalling rises then falls with type. This is the plain hoodie.
Now drag secondary-cue informativeness down toward zero. Watch the high bar overtake the medium and climb to the top: with no other channel to vouch for it, the high type is forced to signal again, and monotonic Spence separation returns. Countersignalling needs a second channel to exist at all.
Back at high informativeness, flip the reader to outsider. The silence that read as high to the insider now reads as low — the outsider can't see the cue, so the countersignal collapses into its opposite. The framed misfire is orientation mismatch.
not signalling as a signal
A signal, in the classical Spence picture, is monotone: more type, more effort, cleanly ordered, so that the amount of signal reads off the amount of quality. Countersignalling breaks the order in the middle. Because the top has a second, independent way to be known, it can afford to let the costly signal go — and once it does, the signal stops meaning "high" and starts meaning "trying to look high," which is a medium's condition. The equilibrium is a genuine fixed point: given that only mediums signal, the high type's best reply is to abstain; given that the high type abstains, the receiver reads silence-plus-cue as high; given that reading, the medium's best reply is still to signal. Everything is consistent, and nothing was assumed — the solver finds it.
The load-bearing condition is the second channel. Turn its informativeness down and the equilibrium changes character: with the cue near useless, the receiver has only the costly signal to go on, silence no longer distinguishes the top from the bottom, and the high type must signal — hardest of all — to be seen. The non-monotonicity is not a law of status; it is a property of environments rich enough in second channels. Abstaining is high-status only when something else is already doing the vouching. Where nothing is, abstaining is just absence.
model
There are three types t ∈ {low, medium, high} drawn from the distribution you set, with underlying standings [2−s, 2, 2+s] where s is the reward spread. Each type chooses a signal effort e on a grid from 0 to 4. Effort is costly, and cheaper for the abler: the cost to type t of effort e is k·e ⁄ θt, with abilities θ = [1, 2, 2.7] — the single-crossing property that lets higher types signal more cheaply.
The secondary cue is an independent binary vouch. Its likelihood is asymmetric and set by informativeness q: a good cue arrives with probability q for the high type, 0.12q for the medium, and never for the low type — so a good cue is specifically evidence of the top, which is what "another channel already vouches for you" means. Each sender observes its own cue before choosing effort. The receiver observes both signal and cue, applies Bayes to get a posterior over types, and pays the posterior-mean standing.
The equilibrium is computed by logit best-response iteration (a quantal-response solve, rationality λ = 6): from the current strategies the receiver's rewards are derived; each type's effort distribution is updated toward its best reply; the map is iterated, damped, to a fixed point — typically in a few dozen steps. Because the strategies keep full support, the receiver's beliefs are always well-defined, with no hand-set off-path assumptions. The tool recomputes the specified game's equilibrium whenever you drag a control.
the act ↔ the model
| The act | The model |
|---|---|
| a type | true standing or skill — the low, medium, or high you actually are, before anyone reads you. |
| signal effort | the costly marker: the flashed credential, the logo, the effort you visibly pour in to be seen as more. |
| the secondary cue | the other evidence that already vouches — reputation, provenance, a competence too plain to need stating. |
| the equilibrium | who ends up signalling how much, once every type is replying optimally to how they will be read. |
| non-monotonicity | the point: mediums loud, highs silent, lows mute — signalling that rises then falls with type. |
| the outsider | one who cannot read the secondary cue, and so misreads the high type's silence as low. |
how this act misfires
misfire · orientation mismatch
The catalogue's named misfire for this act, and the framed one here. A countersignal works by being read against a second channel — the reputation, the provenance, the room. To anyone a step outside that channel, the mechanism is invisible: quiet does not read as security, it reads as low. The plain hoodie says "nothing to prove" to the insider and "nobody" to the stranger. Flip the reader to outsider in the instrument and watch the same silence invert its meaning. The high type who counter-signals to the wrong audience does not look modest; he disappears.
limit · the model assumes a shared, informative channel
The whole non-monotonic result rests on a secondary cue that both parties can read and that genuinely vouches for the top. Drain its informativeness — the case with no other evidence — and there is nothing for silence to be read against: abstaining becomes simply invisible, the high type is forced back into the costly signal, and monotonic Spence separation returns. The pattern is a property of environments dense enough in reputations, provenances, and shared context. In a cold room, everyone signals, and the middle is no longer alone in shouting.
When another channel already vouches for you, silence outranks the signal the anxious can't withhold.
can you use it?
RECOGNITION — Which is the counter-signal? A: a new hire lists every certification in her email signature. B: a famous professor's site is one plain page with her name. C: an unknown freelancer sends a one-line bio to look confident.
B. Her abstention is read against a second channel — reputation — that already vouches. A is the medium's costly marker; C is silence with no cue behind it, which reads as low.
THE NEAREST NEIGHBOR — The high type stays quiet; so does the low type. One criterion tells a counter-signal from mere absence. What is it?
An informative secondary cue. Silence outranks the signal only when something else does the vouching; drain the cue and abstaining is invisibility, and the high type must shout again.
PRODUCTION — You address a room that already knows your work. Compose the counter-signalling introduction, then check it.
“I tinker with these models a bit.” Yours works if it withholds exactly the marker a medium would flash — and only because the room's prior knowledge carries the proof.