The Return Path

In 2001, Marten Scheffer, Stephen Carpenter, Jonathan Foley, Carl Folke, and Brian Walker published "Catastrophic shifts in ecosystems" in Nature. The paper formalized a pattern that field ecologists had been observing for decades: many ecosystems have two stable states, and the transitions between them are not symmetric. A shallow lake with clear water and rooted plants can absorb increasing phosphorus loading until a critical threshold, at which point it flips to a turbid state dominated by phytoplankton. The transition is abrupt. But reducing phosphorus back to the level that triggered the flip does not restore the lake. Recovery requires lowering the phosphorus far below the degradation threshold. The system must overshoot its own history to return.

Scheffer called this hysteresis. The word comes from the Greek hysterein — to lag behind, to come late. It was coined by James Alfred Ewing in 1881 to describe the behavior of iron in a magnetic field. When you magnetize a piece of iron, the relationship between the applied field and the induced magnetism is not a line. It is a loop. Increasing the field produces one curve. Decreasing it produces a different curve. The material retains residual magnetism — remanence — after the field is removed. To demagnetize it, you must apply a reverse field. The forward path and the return path trace different routes through the same space. The B-H curve is the shape of a system that has been somewhere.

What makes hysteresis different from ordinary path dependence: QWERTY persists because switching costs are high, but if everyone switched simultaneously, the system would work fine at any keyboard layout. The current state depends on history, but the state space itself is symmetric. Hysteresis is asymmetric. The system crosses a threshold going one direction and must cross a different threshold going back. The landscape itself changes shape depending on which side you occupy.

Coral reefs exhibit this. A reef stressed by warming, acidification, or nutrient loading can flip to algal dominance. Once established, the algae produce chemical inhibitors that prevent coral larvae from settling. The larvae that would rebuild the reef cannot reach the substrate because the substrate is occupied by the thing that replaced the reef. Tyre Hughes and colleagues documented this in the Caribbean: reefs that lost coral cover in the 1980s had not recovered thirty years later despite reduced fishing pressure and improved water quality. The original stressor was gone. The secondary barrier — the one the degradation itself created — was not.

The Amazon generates roughly half of its own rainfall through transpiration. Trees pull water from the soil, release it from their leaves, and the water falls as rain downwind, feeding the next rank of trees. Carlos Nobre and colleagues argued in 2016 that deforesting past approximately twenty-five percent of the basin could trigger a transition to savanna. Not because savanna is the basin's natural state, but because the forest needs the forest to remain a forest. Remove enough of it, and the self-generated rainfall drops below the threshold that sustains what remains. The degradation destroys the mechanism that would enable recovery.

The Aral Sea demonstrates the irreversible limit case. Soviet irrigation projects diverted the Amu Darya and Syr Darya rivers beginning in the 1960s. By 2007, the sea had lost ninety percent of its volume. Kazakhstan built the Kok-Aral dam in 2005 to save the North Aral Sea, and partial recovery followed within two years. But the South Aral Sea is functionally gone. The exposed lakebed deposited salt across the surrounding landscape. The salt changed the regional climate. The changed climate reduces the rainfall that would refill the basin. The system destroyed its own return path while crossing the threshold.

Ice ages work this way. Milankovitch cycles — variations in Earth's orbital eccentricity, axial tilt, and precession — modulate the distribution of sunlight across latitudes. But the relationship between orbital forcing and ice volume is conspicuously asymmetric: ice sheets grow slowly over tens of thousands of years and collapse rapidly over a few thousand. The asymmetry comes from the ice itself. Growing ice increases albedo, which reflects more sunlight, which cools the surface, which grows more ice. The ice sheet participates in its own expansion. During deglaciation, the reverse feedback (less ice, less reflection, more warming) is accelerated by additional mechanisms — rising CO₂ from warming oceans, meltwater disrupting thermohaline circulation, isostatic rebound changing ocean geometry. The return path is not the forward path reversed. It is a different path through a landscape that the ice sheet itself reshaped.

Scheffer described a diagnostic. In 2009, he and colleagues showed that systems approaching a tipping point exhibit critical slowing down: they take longer to recover from small perturbations. Poke the system and measure how quickly it returns to equilibrium. If the recovery time is lengthening, the basin is flattening — the attractor is losing its grip. The system wobbles more before it falls. The diagnostic works because it measures not the state but the state's relationship to its own stability. By the time the threshold is crossed, that relationship has already changed.

What hysteresis reveals is not that systems resist returning to their previous state. It is that the previous state no longer exists. The system that degraded modified the conditions that defined its former equilibrium. The lake with algae is not a clear lake plus algae. It is a different system — one in which the feedback loops that maintained clarity have been replaced by feedback loops that maintain turbidity. Recovery is not reversal. It is the construction of a new system that happens to resemble the old one, on a landscape that the old one's collapse reshaped.

The hard drive stores data this way. The bit does not remember what the field is doing. It remembers which direction the field last pushed it. The remanence is not a record of the present state. It is a record of the passage. The system carries the shape of its own crossing.