We performed simulations of ground motions in Mexico City expected for large earthquakes in the Guerrero seismic gap in Mexico. The simulation method uses as empirical Green's functions the accelerograms recorded in Mexico City during four small to moderate earthquakes (8 Feb. 1988, Ms = 5.8; 25 April 1989, Mw = 6.9; 11 May 1990, Mw = 5.5; and 31 May 1990, Mw = 6.0) in the Guerrero gap. Because these events occurred in the Guerrero gap, and have typical thrust mechanisms, the propagation path and site effects can be accurately included in our simulation. Fault rupture patterns derived from the 1985 Michoacan earthquake and source scaling relations appropriate for Mexican subduction zone earthquakes are used. If the Guerrero event is similar to the 1985 Michoacan event, the resulting response spectrum in Mexico City will be approximately twice as large as that of the 1985 Michoacan earthquake at periods longer than 2 sec. At periods shorter than 2 sec, the amplitude will be 2 to 3 times larger than that for the Michoacan earthquake. If the events in the Guerrero seismic gap occur as a sequence of magnitude 7.5 to 7.8 events, as they did in the previous sequence around the turn of the century, the strong motion in Mexico City is estimated to be about half that experienced during the 1985 Michoacan earthquake at periods longer than 2 sec. However, several factors affect this estimate. The magnitude of the possible events has a significant range and, if a rupture sequence is such that it enhances ground-motion amplitude with constructive interference, as occurred during the second half of the Michoacan sequence, some components of the ground motion could be amplified by a factor of 2 to 3. To aid in the interpretation of the simulated motion for purposes of design or hazard assessment, design spectra for the CDAO site in Mexico City are derived from the response spectra of the simulated ground motions.