The characteristic decreased recovery and survival of transfused platelets in nonalloimmunized patients with idiopathic thrombocytopenic purpura (ITP) suggest that plasma antiplatelet autoantibodies (autoAbs) are present in almost all cases. Studies emphasizing reactions of IgG autoAbs with platelet glycoprotein (GP) IIb/IIIa indicate that less than 50% of ITP patients have detectable serum Abs, and that many of these Abs may not be pathogenic because they are directed against epitopes in the cytoplasmic domain of GPIIIa (Fujisawa et al, Blood 77:2207, 1991 and 79:1441, 1992). We evaluated the contribution of Ig classes other than IgG to the overall incidence of serum Abs in 47 patients with chronic ITP and the frequency of reactions with GPs IIb/IIIa, Ib/IX, IV, and Ia/IIa. Abs were further characterized by their reactions with cytosolic or exosolic GP epitopes and their titers and apparent affinities. Using immunobead techniques we found (1) anti- GPs in 85% of sera; (2) IgA and IgG Abs each in 68%, together in 51%; (3) IgM agglutinins in 15%, always with another Ab class; (4) GP Ib/IX, IIb/IIIa, IV, and Ia/IIa targets in 83%, 81%, 38%, and 28% of cases, respectively; (5) 93% of positive sera reactive with more than one GP; but GP IV or Ia/IIa never the sole target; (6) Abs against cytosolic epitopes on one or more of GPs IIIa, Ib alpha, and IIb beta in 66% of sera, always accompanied by Abs against exosolic epitopes of the same or a different GP; (7) autoAbs against cytosolic GP epitopes in 38% of 16 patients recovered from posttransfusion purpura and drug purpura; and (8) evidence that serum ITP Abs, often high-titered, saturate platelets less than alloAbs against the same GPs. Whereas Abs against external GP epitopes are a distinctive marker for ITP in 80% of patients, Abs against internal GP epitopes are likely a secondary phenomenon of platelet destruction and not pathogenic. Anti-GPs against exosolic epitopes were also found in eluates of patients platelets', suggesting that they have pathogenic significance.