Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
High-ability agents are more likely to achieve early success in risky experimentation, but learn faster that their project is not promising. These counteracting effects give rise to a signaling model with double-crossing property. This property tends to induce homogenization of quitting times between types, which in turn leads to some pooling in equilibrium. Low-ability agents may hold out to continue their project for the prospect of pooling with the high type, despite having a negative instantaneous net payoff. A war-of-attrition mechanism causes low-ability agents to quit only gradually over time, and to stop quitting for a period immediately before all agents exit.