This paper presents a database of isoseismal maps and referenced/processed (isoseismal contour areas and equivalent radii) data for Colombia and neighboring regions based on earthquakes occurring between 1766 and 2004. The database is used to develop a model for the conditional exceedence probability of a level of seismic intensity, given distance to a site and the epicentral intensity. The model comprises mixed discrete–continuous joint probability distributions. The conditional distribution of distance for a given intensity or intensity difference is continuous and is represented by a lognormal distribution. The distribution of intensities is discrete and is represented by bimodal Poisson functions. The bimodal representation may be due to wave reflections at the boundary between the lithosphere and the asthenosphere in the region. Bayes theorem is used to integrate the continuous and discrete distributions to provide the conditional exceedence probabilities of intensity, given distance and epicentral intensities. Two separate models are proposed, one for shallow, crustal events with depths less than 40 km and the second for subduction events. The models are applied to Colombia and western Venezuela.