The addition of formalin reduces the reactivity of the extracts when they are tested with antisera prepared against "native" antigens and causes changes in the nature of the precipitin curves. This effect has been pointed out by Horsfall (1934) and by Leone (1953) in their work on the effects of formaldehyde on serum proteins. Their data indicate, however, that even though changes in the immunological characteristics of proteins are brought about by formolization, the proteins retain enough of their specific chemical characteristics to allow consistent differentiation of species by immunological methods. In the tests which I performed, the relative positions of the precipitin curves, whether native or formolized extracts were involved, remained unchanged (Figs. 10, 11). _All data used in interpretation of the serological relationships were obtained from tests in which formolized antigens of equivalent age were used._ Only three antisera were produced against formolized antigens, all others being produced against "native" extracts. The formolized antigens seemed to have a greater antigenicity, in most instances, than did those which were unformolized, and precipitin reactions involving antisera produced against formolized antigens developed higher turbidities. The antisera produced against formolized antigens were equal to but no better than those prepared against "native" extracts in separating the birds tested (Figs. 12, 13). The rabbit is a variable to be considered in serological tests. Two rabbits exposed to the same antigen, under the same conditions, may produce antisera which differ greatly in their capacities to distinguish different antigens. It is logical to assume, therefore, that two rabbits exposed to different antigens may produce antisera which also differ in this respect. This explains the unequal values of reciprocal tests shown in Table 2. Thus, in the test involving the antiserum to the extracts of _Richmondena_, a value of 71 per cent was obtained for _Spiza_ antigen, whereas in the test involving anti-_Spiza_ serum, a value of 98 per cent was obtained for _Richmondena_ antigen. In Table 2, therefore, comparisons may be made only among values for the proteins of birds tested with the same antiserum.