When conjugated with antibodies that bind selectively to tumor cells, natural toxins from plants and bacteria represent a potentially powerful form of cancer therapy. Over the last several years, much thought and research effort has been devoted to optimizing the specificity and activity of these hybrid molecules. Here we review some of the insights that have come from detailed analyses of the so-called "first-generation immunotoxins," and discuss how these insights have prompted ideas for more effective design of "second-generation immunotoxins."