Why does additional schooling raise the wages of those who complete it? For decades the dominant answer was human-capital theory, which holds that education raises earnings because it makes workers more productive: instruction imparts knowledge and skills that employers value and reward. On this account the wage premium enjoyed by graduates measures, however imperfectly, a genuine enhancement of their capacities. A rival account, the signaling model, accepts that education raises earnings but denies that it must do so by raising productivity. On this view, schooling functions partly as a sorting device. Because completing a demanding course of study is easier for the already-able - the diligent, the intelligent, the conscientious - a credential can reliably certify pre-existing traits that employers cannot observe directly. Employers pay graduates more not because school made them more capable but because the diploma testifies to capabilities they possessed beforehand. The classroom, in this telling, may teach little that the workplace uses; its function is to separate wheat from chaff. The two theories are notoriously difficult to distinguish empirically, because both predict the central fact to be explained - that education and earnings move together. Distinguishing them requires evidence about mechanism. One suggestive pattern is the so-called sheepskin effect: workers who complete a degree earn markedly more than those who accumulate nearly the same coursework without finishing. If wages tracked skill acquired, the final credit before graduation should matter little; that it matters greatly suggests the credential itself, not the learning it nominally represents, commands the premium. Human-capital theorists respond that the final year may be unusually productive, or that graduation certifies the completion of skills that only cohere at the end. Neither theory, in its pure form, is likely correct, and few of their proponents hold one. The interesting disputes concern proportions: how much of the premium reflects skill genuinely built, how much mere certification. The distinction is not academic. If education is largely productive, then expanding access enriches the workforce and, plausibly, the society. If it is largely a signal, then subsidizing ever-more schooling may chiefly intensify a credential race - raising the educational threshold for a given job without raising the competence brought to it, and imposing on later entrants the cost of credentials that earlier ones did not need. Even this framing, however, understates the entanglement. A credential that begins as a pure signal can become productive if employers, having sorted workers by it, then assign the credentialed to roles where they accumulate valuable experience. Conversely, genuinely useful instruction can lose its productive rationale and persist as ritual once employers come to expect it. The mechanisms are not fixed properties of education but evolving products of how institutions use it. To ask whether education signals or builds, then, may be to pose as a permanent dichotomy what is better understood as a variable and historically contingent mixture.