Escalation of commitment beats sunk cost: fMRI picks the winner

Economists crowned the wrong villain. For decades, the sunk cost fallacy got the headlines, the textbooks, and the TED talks. But when neuroscientists at Oxford put 30 people into an fMRI scanner and tracked every decision through 1,247 trials, the real culprit turned out to be its quieter cousin: escalation of commitment. And it is roughly 3.4x more destructive to your choices than sunk costs alone, because it does not wait for money to be spent. It hijacks your attention before the cost even registers.

This distinction matters more than it sounds. If the sunk cost fallacy is "I already paid, so I'll stay," escalation of commitment is "I'm already moving, so I'll keep moving, harder." One is about loss. The other is about identity. And identity is what Netflix, Duolingo, and your trading app are actually selling you.

The brain region that makes you double down

The 2024 study by Eleanor Holton and colleagues, published in Nature Human Behaviour, found that the ventromedial prefrontal cortex (vmPFC, the brain region behind goal pursuit and self-relevant thinking) lights up every time you get closer to a goal, whether that goal is worth finishing or not. It narrows your attention. It dims the appeal of alternatives. It literally makes the next episode, the next trade, the next lesson feel more important than it is.

Here is the twist that destroys the classic sunk cost story: patients with vmPFC damage performed better in the exact same task. They quit failing goals faster. They noticed better options. Their brains, lacking the persistence circuitry, did not suffer the bias that healthy brains display by default.

This is not a willpower problem. It is wiring.

Sunk cost is the symptom, escalation is the mechanism

A separate fMRI study from 2013 in Brain Research mapped the two phenomena onto completely different neural pathways. Sunk costs activated lateral frontal and parietal regions linked to risk evaluation. Escalation triggered striatal reward circuits, the same ones that fire during slot machine play. The researchers found zero overlap between the two systems. That is why strategies built for one almost never fix the other. You can intellectually accept a loss and still feel pulled to keep going, because the pull is not in the accounting part of your brain. It is in the wanting part.

A meta-analysis by Xu and colleagues (776 participants across four experiments) sealed it: escalation is driven by negative feedback, not by the sum of past investment. The worse a trajectory feels, the harder people press forward. The original sunk cost framing misses this entirely.

How Netflix weaponizes the same circuit

Netflix figured this out before the neuroscientists did. Their autoplay countdown, the "continue watching" rail, the 80 percent of viewing that flows from algorithmic recommendations: all of it is calibrated to one behavior. Not "you already invested time," but "you are already in motion."

The 15-second countdown before the next episode is a masterpiece of escalation design. It does not ask you to choose to continue. It asks you to choose to stop. That is a completely different decision for the vmPFC, and one it is wired to lose. Duolingo's streak system, Robinhood's confetti, your trading app's push notifications after a bad session: same mechanism, different wrapper. They are not showing you sunk costs. They are protecting your sense of forward motion.

This is the same psychological trap that silently drains millions from businesses and the one behind the decision fatigue breakdown at choice #227.

What actually works (and what doesn't)

Most sunk cost "cures" tell you to ignore what you've spent. Fine. It does not help, because escalation is not about spending. The research points to three things that actually move the needle:

1. Interrupt the motion, not the math. Close the app mid-episode, exit the trade without reviewing P&L, stop the streak before it finishes. The break in momentum is the intervention, not the accounting.
2. Swap identity frames. "I am someone who finishes" predicts escalation. "I am someone who reallocates" predicts healthier quitting. Holton's data showed vmPFC activity tracked identity-relevant goals most strongly.
3. Set pre-commitment exits. Decide the quitting condition before you start, when the vmPFC is not yet invested. By episode four, it is too late to ask yourself whether you want to be watching.

The lesson is not that sunk cost is wrong. It is that sunk cost is the surface. Escalation of commitment is the current, and almost every product competing for your time has already built a paddle.

---

Related Reading:

• I tracked every trade for 365 days: my brain lost me $4,200
• You spend $219/month on subscriptions: your brain sees $86