We have further analysed the roles of mesoderm induction and dorsalization in the formation of a regionally specified mesoderm in early embryos of Xenopus laevis. First, we have examined the regional specificity of mesoderm induction by isolating single blastomeres from the vegetalmost tier of the 32-cell embryo and combining each with a lineage-labelled (FDA) animal blastomere tier. Whereas dorsovegetal (D1) blastomeres induce 'dorsal-type' mesoderm (notochord and muscle), laterovegetal and ventrovegetal blastomeres (D2-4) induce either 'intermediate-type' (muscle, mesothelium, mesenchyme and blood) or 'ventral-type' (mesothelium, mesenchyme and blood) mesoderm. No significant difference in inductive specificity between blastomeres D2, 3 and 4 could be detected. We also show that laterovegetal and ventrovegetal blastomeres from early cleavage stages can have a dorsal inductive potency partially activated by operative procedures, resulting in the induction of intermediate-type mesoderm. Second, we have determined the state of specification of ventral blastomeres by isolating and culturing them in vitro between the 4-cell stage and the early gastrula stage. The majority of isolates from the ventral half of the embryo gave extreme ventral types of differentiation at all stages tested. Although a minority of cases formed intermediate-type and dorsal-type mesoderms we believe these to result from either errors in our assessment of the prospective DV axis or from an enhancement, provoked by microsurgery, of some dorsal inductive specificity. The results of induction and isolation experiments suggest that only two states of specification exist in the mesoderm of the pregastrula embryo, a dorsal type and a ventral type. Finally we have made a comprehensive series of combinations between different regions of the marginal zone using FDA to distinguish the components. We show that, in combination with dorsal-type mesoderm, ventral-type mesoderm becomes dorsalized to the level of intermediate-type mesoderm. Dorsal-type mesoderm is not ventralized in these combinations. Dorsalizing activity is confined to a restricted sector of the dorsal marginal zone, it is wider than the prospective notochord and seems to be graded from a high point at the dorsal midline. The results of these experiments strengthen the case for the three-signal model proposed previously, i.e. dorsal and ventral mesoderm inductions followed by dorsalization, as the simplest explanation capable of accounting for regional specification within the mesoderm of early Xenopus embryos.