In two experiments carried out in Guadeloupe, barriers were used to reduce the entry of the virus vector Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae)to tomato plots. The barriers erected around the crop were of insect-proof cloth fences (<50 mesh), 1.5 m in height, in the first experiment with a deltamethrintreated, insect-attracting strip facing inwards, and, in the second, with the barrier but no insecticide-treated strip. 2 A mathematical model of epidemic development was fitted to the symptom data from the treated and control (unprotected) tomato plots. There were two viruses present, tomato yellow leaf curl and potato yellow mosaic; specific detection confirmed that symptoms gave an accurate indication of infection and that the two virus diseases had similar progress curves. 3 Parameter estimates obtained by model-fitting suggested that the barriers reduced vector immigration by approximately 12-fold but that B. tabaci retention within the plots was also increased slightly despite the mortality caused by the insecticide-treated strips. Disease establishment was delayed by approximately 2 weeks. The results obtained in the second experiment involving barriers deployed without insecticide-treated strips could be explained by a large increase in B. tabaci retention within the barriers resulting in more rapid virus disease progress than in controls. The results of mathematical modelling indicate that partial insect barriers can be worse than none because sufficient whiteflies can enter to establish a population and, at the same time, large numbers are retained in the barrier plot, with the net effect being a more rapid population increase than in the absence of barriers.