To examine the use of acetone- or ethanol-fixed frozen tissue sections as the "gold standard" for immunohistochemical analysis, we evaluated frozen sections with various conditions of fixation and antigen retrieval (AR). Fresh human tissues were frozen in OCT. An adjacent tissue block was fixed in 10% neutral buffered formalin (NBF) and paraffin embedded (FFPE). Frozen sections were fixed by 6 protocols: acetone, ethanol, NBF (2 durations), and NBF + calcium chloride (2 durations). AR was used for all NBF-fixed sections. More than half of the antibodies (16/26 [62%]) showed immunohistochemical results indistinguishable between acetone- and NBF-fixed sections; 8 (31%) showed better immunohistochemical signals following NBF and AR; 2 gave better immunohistochemical results for acetone-fixed sections. Most cytoplasmic proteins (10/13) showed comparable immunohistochemical signals between acetone- and NBF-fixed sections. For nuclear proteins, NBF-fixed sections gave better immunohistochemical signals than did acetone-fixed sections. In most cases, NBF yielded stronger signals with less background and better morphology. The data do not support the use of acetone-fixed frozen tissue sections as the gold standard for immunohistochemical analysis. In evaluating new antibodies, a combination of acetone- and NBF-fixed frozen sections should be used, although in practice, FFPE tissue sections may serve as the standard for most antigens for immunohistochemical analysis.