Abstract: Class action law is built on a model that assumes a large group of individuals have similar legal claims of such small value that no one of them has the incentive or ability to litigate alone. Rule 23 resolves that collective action problem by enabling one class member to represent the group, with a common fund fee award sharing the costs across the class. The Constitution guarantees class members the options of opting out (exit) or objecting (voice), but given the small stakes, most do nothing (loyalty). While elegant, this model does not capture the reality of all class suits. In many cases, some class members have significant enough legal claims that they are capable of litigating alone. The group dynamics accordingly change, with everything turning on the decisions of the large claimants: in some securities cases, they step forward to perform a monitoring function (voice) and often they simply remain passive (loyalty), but the central question is whether they will opt out and litigate separately in the hopes of maximizing their recovery (exit). The risk that they might deters defendants from settling the class’s small claims, lest they then have to litigate the large claimants’ valuable claims. But the risk simultaneously creates an opportunity: if the class could present a united front, a defendant would likely pay a premium to settle the whole package of claims. Heterogeneous classes can therefore suffer a problem akin to a prisoner’s dilemma: every class member might be best off if they could work together, but lacking a mechanism to do so, coordination costs render the option elusive. The tragedy of this commons is that, built on a different template, class action law offers class members only the three options of exit, voice, or loyalty. In this Article, we offer heterogeneous class members a fourth option: cooperation. Our proposed mechanism for harnessing claimants’ cooperative instincts is a new form of class certification that we call “negotiation class certification.” Under this approach, class members would work together to generate a metric for distributing a lump sum settlement across the class. They would then ask the court to certify a “negotiation class” and to direct notice to the class members informing them that counsel will negotiate a lump sum settlement and that, if achieved, the lump sum amount would be put to a vote, with a supermajority vote binding the class; the notice would also explain the distributional metric. Any class member that did not want to bind itself to either the distributional metric or the supermajority voting process could opt out. By establishing the contours of the class prior to settlement discussions, negotiation class certification would provide the defendant with a precise sense of the scope of finality a settlement would produce, hence encouraging a fulsome offer. The proposal is a novel use of Rule 23, but it is, in many ways, a less ambitious one than certification of a settlement class, although the latter approach, controversial at its inception, has been a “stock device” in class action practice for nearly a quarter century. And while novel, negotiation class certification is consistent with the requirements of both Rule 23 and the Constitution. Indeed, engaging large class members in the settlement negotiation process ex ante improves on a system that delegates that authority to agents and involves the class only ex post.