A theory for data‐aided equalization and cancellation in digital data transmission over dually polarized fading radio channels is presented. The present theory generalizes and extends previous work by admitting decision feedback structures with finite‐tap transversal filter implementations. Subject to the assumption that some past and/or future data symbols are correctly detected, formulas and algorithms for evaluating the least mean‐square error for different structures are presented. In a sequence of curves we evaluate and compare the performance of various structures for a particular propagation model and several fading events. We find improvement in performance for decision feedback over linear equalization. More importantly, we discovered that in this application, as in the single‐channel transmission case, decision feedback/canceler structures are much less sensitive to timing phase than linear equalizers.