Simultaneous consideration of source, path, and site effects on ground motion during the Michoacán earthquake of 1985 allows us to draw coherent conclusions regarding the roles played for the disaster in Mexico City by the rupture process, the mode of propagation of the waves between the epicentral zone and Mexico City, and the local amplification. In contrast to the horizontal component which showed dramatic amplification for the 2 to 3 sec motion at lake sediment sites, we observe almost identical vertical displacement seismograms containing ripples with 2 to 3 sec period throughout the Mexico City valley whether the recording site is on the lake sediments or on hard rock. We, therefore, conclude that the 2 to 3 sec motion responsible for the destruction of Mexico City was present in the incident field. After performing a phase analysis, we interpret the signal as the superposition of long-period Rayleigh waves and short-period Lg with a dominant period of about 3 sec. The analysis of the teleseismic records indicates that the radiation of this event is enhanced for waves around the 3 sec period. Except in the case of stations for which an anomalous path effect is suspected, the records present ripples appearing a few seconds after the beginning of the signal. The characteristics of near-fault records show that the rupture process consists of the growth of a smooth crack. The numerical simulation indicates that the 3 sec period ripples can be explained by a series of changes of the rupture front velocity. We examine two alternative source models associated with different crustal models to explain the characteristics of the vertical displacements recorded in Mexico City. Our preferred model attributes the cause of the enhanced 3 sec motion to the irregularity in the rupture propagation in addition to the effect of the local conditions in Mexico City. This interpretation leads to a very coherent scenario of what happened from the start of the failure on the fault up to the destruction in Mexico City. This example illustrates the need to consider simultaneously source, path, and site effects in order to understand strong ground motions.