"Publication bias" has three facets: (1) bias perceived by disappointed authors, (2) bias that journal policy may introduce, and (3) bias intrinsic in design and interpretation of the work itself. The third type, though a target of peer review, is not considered here, and the first type is more often imagined than real. However, general journals have to adopt policies on priorities that an outsider may see as bias, in the broadest sense. Opportunities for bias exist (18 varieties are listed here), but more objective evaluation is required before journals need to alter their peer review practices. In terms of work load and financial considerations, the price of some proposed correctives is high. Journals should monitor refereeing systems and allow appeals, but the case for policing systems--"blinding" referees (and, logically, editors), introducing tight codes of practice, and seeking solemn declarations of integrity, for example--needs more hard evidence.