I document a dramatic increase in the importance of two types of variation in Treasury bill yields beginning in the early 1980s. The first is idiosyncratic variation in individual short-maturity (less than three months) bill yields. The second is a common component in Treasury bill yields that is not shared by yields on other instruments, such as short-maturity privately-issued instruments or longer-maturity Treasury notes and bonds. Some evidence suggests the first type reflects increased market segmentation. These results have important implications for the calibration and testing of no-arbitrage term structure models and interpreting tests of the expectations hypothesis.