To assess the variation in antibacterial activity of honey a survey was carried out on 345 samples of unpasteurized honey obtained from commercial apiarists throughout New Zealand. Most of the honeys were considered to be monofloral, from 26 different floral sources. The honeys were tested against Staphylococcus aureus in an agar well diffusion assay, with reference to phenol as a standard. Antibacterial activity was found to range from the equivalent of less than 2% (w/v) phenol to 58% (w/v) phenol, with a median of 13.6 and a standard deviation of 12.5. Neither the age of the honey samples nor whether they had been processed by the apiarist was associated with lower activity. However, the difference between floral sources in the antibacterial activity was very highly significant. Kanuka (Kunzea ericoides (A. Rich.) J. Thompson. Family: Myrtaceae), manuka (Leptospermum scoparium J. R. et G. Forst. Family: Myrtaceae), ling heather (Calluna vulgaris (L.) Hull. Family: Ericaceae) and kamahi (Weinmannia racemosa Linn. f. Family: Cunoniaceae) were shown to be sources likely to give honey with high antibacterial activity. When antibacterial activity was assayed with catalase added to remove hydrogen peroxide, most of the honeys showed no detectable antibacterial activity. Only manuka and vipers bugloss (Echium vulgare L. Family: Boraginaceae) honeys showed this type of activity in a significant proportion of the samples. The high antibacterial activity of manuka honey was in many cases due entirely to this non-peroxide component.