In a controlled field setting, in which the majority of people in our sample lose more than £90,000 ($120,000), we examine how human beings respond to major financial losses. University ethics boards would not allow this kind of huge-loss phenomenon to be studied with normal social-science experiments. Yet the scientific and practical issues at stake are unusually important ones. In our setting, individuals are handed £100,000 in cash. They then have to make risky decisions. Facing a sequence of seven questions, individuals are required to distribute their cash endowment over a set of possible answers. Participants lose any cash placed on a wrong answer. We find evidence of risk reduction after people suffer a loss in the previous decision round. A prior financial loss of £10,000 is estimated to increase the propensity to fully diversify by 6 percentage points. In terms of proportional losses, a loss of 50% or more of the remaining cash endowment increases diversification rates by approximately 13 percentage points. The fixed-effects panel data estimates are robust to the remaining cash endowment, previous diversification strategy, relative difficulty of questions, the ability level of participants, and other personal traits. The findings support a prospect theory-based model with a coefficient of loss aversion that is increasing in past losses. Our study appears to be the first to be able to calculate systematically how human beings react to enormous and unrecoverable financial losses.