Usually, main lessons in dam engineering were dressed up from observations after earthquakes. The JCOLD-CFBR exchanges show that other major lessons are learn from the seismic measurements recorded on dams during the earthquakes. Japan is the first country over the world which has installed and managed a huge net of seismographs on dams. By this way, JCOLD has a deep insight in seismic dam performance. Thanks to the release of N°2 seismic data base in the ICOLD congress of Montreal (2003) and of N°3 data base in the ICOLD congress of Stavanger (2015) after the great Tohoku earthquake (M = 9), the qualification of any seismic analysis, based on comparison of modeled results with measured data, can be assessed before using it in dam safety assessment. The discrepancies of such comparisons show the weaknesses of modeling and the needed improvements. These data provide the natural periods of the instrumented dams. Correlations, giving the natural period versus height and type of dams, are processed. However, calibration of elastic modulus of dams without accelerometers could be carried out with measurements and interpretation of micro-tremors from ambient vibrations. More research is needed to understand and calibrate the damping. Modulus reduction curve was deduced from measured accelerations at Aratozawa dam. The weakest dams are affected by a strength reduction increasing with the energy of the earthquake. This strength reduction has been ignored by most of simplified methods; either it was not implemented or it was managed with many problems in the finite element method. Tatsuoka in a companion paper proposed some solutions to simulate this loss of strength. Practitioners are invited to take account this loss of strength in simplified or sophisticated analyses.