The question of how to tax married couples has remained controversial since the adoption of the federal income tax over a century ago. The appropriate answer to this question depends in part on how sensitive couples are to taxes when deciding to marry. Yet we know surprisingly little about how taxes shape couples' marriage choices. This Article begins to fill that gap using a natural experiment generated by the halting shift in how the income tax treated married couples in the mid-Twentieth Century. The system moved from taxing married couples as two individuals--in which case marriage largely did not affect taxes--to taxing married couples on their joint income. At the time of this shift, joint taxation lowered couples' taxes upon marriage. The change to joint taxation, however, came later to some states, creating a natural experiment to study its impact on marriage rates. This Article shows that annual marriage rates increased by 9% in the relevant states after the introduction of joint taxation made marriage tax-advantaged, with affected men marrying 3 to 5 months sooner on average. This suggests that at any given time during this period there were tens or hundreds of thousands of married couples in the United States who would not have been married if not for the tax incentives. Couples appear to have been unexpectedly responsive to the tax changes given that unmarried cohabitation was not acceptable under the social mores of the day. If anything, Americans today are likely more sensitive to taxes when deciding whether and when to marry, suggesting that joint taxation continues to affect marriage decisions today. This in turn strengthens the case for returning to individual taxation of marriage if the goal is to avoid inefficiently distorting people's marriage decisions. By contrast, if the government wishes to encourage marriage, the results imply that using the tax code may be effective under some circumstances, but further analysis suggests that joint taxation remains a poor choice for doing so.