Present methods for accelerometer calibration include small‐amplitude sinusoidal drive, large‐amplitude sinusoidal drive, and pulse‐integration (ballistic pendulum) methods. The first of these is of limited use because it fails to reveal nonlinear effects and pulse saturation. The second is difficult to accomplish except at a limited number of frequencies, while the third gives very little high‐frequency information. A method of direct calibration, using bonded wire strain gauges as a standard, is described. The method is valid over a very wide frequency range, up to 40 kc with present instrumentation. Several examples of the use of the method are discussed. Results show that the method is an easy one to use and that it quickly reveals any deficiencies of accelerometers and their amplifiers, if present, e.g., “droop” resulting from insufficient low‐frequency response, overloading of cathode follower, loss of high‐frequency detail caused by too low cutoff of accelerometer filter, hash caused by accelerometer resonance, etc. Limitations of the method and evaluation of accuracy also are discussed. This method has been found to give a complete picture of the actual operation of the accelerometer and associated circuits quickly and accurately, and the results are easily interpreted.