Regression equations modeling gas chromatographic retention behavior of polychlorinated dibenzodioxins were generated for non-polar, moderately polar, and polar stationary phases. Topological, electronic and geometrical structure-based descriptors were calculated and related to the molecules' retention behavior by using a variety of linear regression techniques. Descriptors focusing on the bridgehead carbons were also evaluated. Model validation techniques and outlier detection methods were integrated into the analyses. Two models with R > 0.999 and relative standard error < 1% were developed for the non-polar columns DB-5 and SP-2100. A moderately polar column, OV-1701, yielded a three-descriptor model with R = 0.996 and relative standard error = 1.1%, and a polar column, SP-2330, yielded a three-descriptor model with R = 0.988 and relative standard error = 1.0%.