We have estimated the effect of omitting antagonism of neuromuscular block on postoperative nausea and vomiting. A systematic search (MEDLINE, EMBASE, Biological Abstracts, Cochrane library, reference lists and hand searching; no language restriction, up to March 1998) was performed for relevant randomized controlled trials. In eight studies (1134 patients), antagonism with neostigmine or edrophonium was compared with spontaneous recovery after general anesthesia with pancuronium, vecuronium, mivacurium or tubocurarine. On combining neostigmine data, there was no evidence of an antiemetic effect when it was omitted. However, the highest incidence of emesis with neostigmine 1.5 mg was lower than the lowest incidence of emesis with 2.5 mg. Numbers-needed-to-treat to prevent emesis by omitting neostigmine compared with using it were consistently negative with 1.5 mg, and consistently positive (3-6) with 2.5 mg. There was a lack of evidence for edrophonium. In two studies, three patients with spontaneous recovery after mivacurium or vecuronium needed rescue anticholinesterase drugs because of clinically relevant muscle weakness (number-needed-to-harm, 30). Omitting neostigmine may have a clinically relevant antiemetic effect when high doses are used. Omitting antagonism, however, introduces a non-negligent risk of residual paralysis even with short-acting neuromuscular blocking agents.