The chemical composition of ferromanganese nodules from the three nodule-bearing MANOP sites in the Pacific can be accounted for in a qualitative way by variable contributions of distinct accretionary processes. These accretionary modes are: 1. (1) hydrogenous, i.e., direct precipitation or accumulation of colloidal metal oxides in seawater, 2. (2) oxic diagenesis which refers to a variety of ferromanganese accretion processes occurring in oxic sediments; and 3. (3) suboxic diagenesis which results from reduction of Mn+4 by oxidation of organic matter in the sediments. Geochemical evidence suggests processes (1) and (2) occur at all three MANOP nodule-bearing sites, and process (3) occurs only at the hemipelagic site, H, which underlies the relatively productive waters of the eastern tropical Pacific. A normative model quantitatively accounts for the variability observed in nearly all elements. Zn and Na, however, are not well explained by the three end-member model, and we suggest that an additional accretionary process results in greater variability in the abundances of these elements. Variable contributions from the three accretionary processes result in distinct top-bottom compositional differences at the three sites. Nodule tops from H are enriched in Ni, Cu, and Zn, instead of the more typical enrichments of these elements in nodule bottoms. In addition, elemental correlations typical of most pelagic nodules are reversed at site H. The three accretionary processes result in distinct mineralogies. Hydrogenous precipitation produces ??MnO2. Oxic diagenesis, however, produces Cu-Ni-rich todorokite, and suboxic diagenesis results in an unstable todorokite which transforms to a 7 A?? phase ("birnessite") upon dehydration. The presence of Cu and Ni as charge-balancing cations influence the stability of the todorokite structure. In the bottoms of H nodules, which accrete dominantly by suboxic diagenesis, Na+ and possibly Mn+2 provide much of the charge balance for the todorokite structure. Limited growth rate data for H nodules suggest suboxic accretion is the fastest of the three processes, with rates at least 200 mm/106 yr. Oxic accretion is probably 10 times slower and hydrogenous 100 times slower. Since these rates predict more suboxic component in bulk nodules than is calculated by the normative analysis, we propose that suboxic accretion is a non-steady-state process. Variations in surface water productivity cause pulses of particulate flux to the sea floor which result in transient Mn reduction in the surface sediments and reprecipitation on nodule surfaces. ?? 1984.