A British engineer is designing gas masks.
His problem is not the amount of charcoal in the filter. It is whether the gas finds a path through.
One grain absorbs, another lets it pass; between them, a network of open and closed pores. The question is not how many pores are open. It is: do the open pores link up from one edge to the other?
From this question, raised in the mid-1950s, Broadbent and Hammersley (1957) draw a model of great austerity. Take a regular lattice. Each link is open with probability p, closed otherwise, independently of the others. A fluid is injected at one point. Does it reach infinity?
The answer does not vary continuously with p. It tips over.
Below a critical value, the fluid stays trapped in finite clusters; every path dies at short range. Above it, a single cluster spans the entire lattice. Between the two, nothing: the transition is sharp. For the bonds of the square lattice, Kesten (1980) proved that this threshold equals exactly one half.
The threshold is not a property of any link.
No open pore decides passage. No link is, by itself, a path. The spanning cluster is a global object: it exists or it does not, and that existence cannot be read anywhere in the inventory of links taken one by one. A link becomes load-bearing only with respect to the whole that runs through it. The whole decides what each part was.
The threshold was inscribed in the structure before the first link. It does not depend on the order in which the pores opened, but on the geometry of the lattice. The critical value precedes any particular realization. It is a condition the lattice carries, independently of what fills it.
The same tipping reappears wherever elements connect at random.
Erdős and Rényi (1960) observe it in random graphs. As long as each node has, on average, fewer than one link, the graph remains an archipelago of small components. The moment the average crosses one, a giant component absorbs a finite fraction of the whole. The network passes from dust to continent with no intermediate state.
An epidemic obeys the same threshold. Kermack and McKendrick (1927) formulate it: below a critical reproduction rate, outbreaks die out in short chains; above it, the infection runs through the population. Newman (2002) shows that this tipping is, mathematically, a percolation on the contact network. The disease does not decide. Connectivity decides.
A gel obeys the same threshold. Flory and Stockmayer describe setting: as long as the molecules bind into finite clusters, the medium stays liquid; at the gel point, a single molecule suddenly links the whole vessel. Egg white that sets, rubber that vulcanizes, cross this point. Before it, a liquid; after it, a solid. The boundary is a matter of connection, not of quantity.
A rumor, a panic, a norm obey the same threshold. Granovetter (1978) models collective action as a cascade of linked triggers. Below a certain density of coupling, the movement halts at a few individuals; above it, it runs through the group. What separates the failed protest from the general one is not the intensity of the grievance. It is the structure of the links that carry it.
In all these cases, the threshold separates two regimes with no gentle transition. Below, the future is fragmented: a multitude of short, disjoint fates that do not communicate. Above, a single connected fate runs through the whole. And from inside a node, nothing distinguishes the two regimes. The local link is identical. Only the whole reveals which side one is on.
Doctrine
Connectivity is not a sum of links. It is a threshold.
No element carries passage on its own. A system's capacity to be traversed is a global property, fixed by its structure before any particular link exists. The threshold precedes the realization. It decides retroactively what each link will have been: inert or load-bearing, lost in a finite cluster or a link in the path that connects everything.
Open vector
We manage systems link by link, as if each connection counted for itself.
But a society's capacity to transmit a signal, an alarm, a refusal, a solidarity, is decided by a threshold we do not see. Two societies, one just below, the other just above, are locally indistinguishable: same individuals, same links, same apparent density. One smothers what passes through it; the other propagates it entirely.
The question, then, is not: how many links have we woven? The question is: which side of the threshold are we on — and can we ever know it from the inside?
