Hafele and Keating, Nature 227 (1970), pg 270 (proposal).
Science Vol. 177 pg 166–170 (1972) (experiment).
They flew atomic clocks on commercial airliners around the world in both directions, and compared the time elapsed on the airborne clocks with the time elapsed on an earthbound clock (USNO). Their eastbound clock lost 59 ns on the USNO clock; their westbound clock gained 273 ns; these agree with GR predictions to well within their experimental resolution and uncertainties (which total about 25 ns). By using four cesium-beam atomic clocks they greatly reduced their systematic errors due to clock drift.
Criticised in: A. G. Kelly, “Reliability of Relativistic Effect Tests on Airborne Clocks”, Inst. Engineers Ireland Monograph No. 3 (February 1996), http://www.cartesio-episteme.net/H&KPaper.htm. His criticism does not stand up, as he does not understand the properties of the atomic clocks and the way the four clocks were reduced to a single “paper” clock. The simple averages he advocates are not nearly as accurate as the paper clock used in the final paper—that was the whole point of flying four clocks (they call this “correlated rate change”; this technique is used by all standards organizations today to minimize the deficiencies of atomic clocks).