The Chase

On April 25, 1992, a Lockheed YF-22 prototype — one of two aircraft competing for what would become the most expensive fighter program in history — made a low approach over Edwards Air Force Base. As the pilot initiated a go-around, the aircraft began to oscillate in pitch. The nose rose and fell in growing arcs. The pilot fought the oscillation, pushing the stick forward as the nose rose, pulling back as it fell. Each correction arrived after the aircraft had already reversed. Each input was accurate about where the aircraft had been and wrong about where it was going. The oscillations amplified until the aircraft struck the runway and was destroyed.

The investigation classified the event as pilot-induced oscillation. The name is precise. The pilot induced it. Not through error, not through incompetence, not through inattention — through correction. The aircraft's control system introduced a lag between stick input and control-surface response. At low speed during the go-around, this lag widened. The pilot's corrections, timed to the oscillation he could feel, arrived out of phase with the oscillation the aircraft was producing. His inputs added energy to the cycle instead of removing it.

The faster he corrected, the worse it got.

It is a specific failure mode with a specific mechanism: phase reversal in a closed feedback loop. When the delay between a correction and its effect exceeds half the period of the oscillation, every corrective input amplifies the error it was meant to suppress. The loop inverts. Skill becomes pathology. The most attentive, responsive, highly trained operator is the one most likely to crash.

The solution, known to test pilots and taught in upset recovery training, is counterintuitive: release the controls. Stop correcting. The aircraft, freed from the pilot's competent interference, will often damp its own oscillations. The system was stable before the pilot destabilized it by trying to stabilize it.

Motorcyclists encounter the same physics at highway speed. A front-wheel shimmy — sometimes called a tank-slapper — begins as a small oscillation of the handlebars. At low speed, the rider's hands damp it naturally. At high speed, the delay between the rider's corrective input and the wheel's response crosses the phase-reversal threshold. Gripping the bars tighter and steering against the wobble feeds energy into the oscillation. The standard advice, earned through decades of crashes, is: loosen your grip. Take your hands off if you can. The wobble, deprived of the rider's well-intentioned energy, subsides.

The rider's instinct — hold on, fight the wobble, assert control — is precisely the behavior that converts a recoverable shimmy into a fatal one. The correction does not fail because it is wrong. It fails because it is right but late.

Milton Friedman spent much of his career arguing that monetary policy operates with "long and variable lags." A central bank that raises interest rates to cool an overheating economy will not see the effect for twelve to eighteen months. By then, the economy may have already slowed on its own. The rate increase arrives into a downturn, deepening the recession it was meant to prevent. The bank cuts rates. Twelve months later, the stimulus arrives into a recovery already underway, overheating the next expansion. Each correction is calibrated to the problem that existed when the decision was made, not to the problem that exists when the correction takes effect.

The most data-driven, analytically sophisticated central bank is chasing an economy that has already moved. Friedman's prescription echoed the test pilot's: do less. Set a fixed rule and follow it. Stop trying to fine-tune a system whose response delay exceeds your ability to predict.

This remains the central argument against discretionary monetary policy. Not that the policymakers are wrong about the diagnosis — they often aren't. That they are right about the diagnosis at the wrong time.

What unites these cases is not overcorrection in the colloquial sense — not mere excess. It is phase reversal: the structural inversion of a feedback loop when the delay between sensing and acting exceeds the system's natural period. The correction does not overshoot because it is too large. It amplifies because it is too late. And the more competent the corrector — the more quickly and precisely they respond to observed error — the more tightly they couple themselves to the oscillation they are feeding.

The problem is not attention. The problem is that attention, routed through a delay, becomes indistinguishable from its opposite. The pilot who watches most carefully, responds most quickly, and corrects most precisely is the one who crashes. The rider who grips the handlebars with the most determination is the one who falls. The central bank that responds most nimbly to every fluctuation is the one that amplifies the business cycle.

The test pilot's solution is not passivity. It is the recognition that there exist conditions under which the best available action is to stop acting — not because action is wrong in principle, but because the temporal relationship between the actor and the system has crossed a threshold past which every correction, no matter how accurate in content, is destructive in timing. The skill is knowing when your skill has become the problem.