Experimental evaluation of the effectiveness of resistance management tactics is vital to help provide guidelines for the deployment of transgenic insecticidal crops. Transgenic broccoli expressing a Cry1Ac gene of Bacillus thuringiensis (Bt) and the diamondback moth, Plutella xylostella (L.), were used in greenhouse tests to evaluate the influence of size and placement of nontransgenic refuge plants on changes in resistance allele frequency and pest population growth. In the first test with an initial Cry1Ac-resistance (R) allele frequency of 0.007, P. xylostella were introduced into cages with the following treatments: 0, 3.3, 10, 20, and 100% refuge plants. Results after four generations showed that resistance could be delayed by increasing the proportion of refuge plants in the cage. Population growth was also influenced by refuge size with the highest populations occurring in treatments that had either no refuge plants or all refuge plants. In the second test, we evaluated the effect of refuge placement by comparing 20% separate and 20% mixed refuges. P. xylostella with an initial frequency of resistant alleles at 0.0125 were introduced into cages and allowed to cycle; later generations were evaluated for resistance and population growth. Separating the refuge had a pronounced effect on delaying resistance and slowing establishment of resistant larvae on Bt plants. Combining information from both trials, we found a strong negative correlation between the number of larvae on Bt plants and the mortality of the population in leaf dip bioassays. Results from larval movement studies showed that separate refuges delayed resistance better than mixed refuges because they conserved relatively more susceptible alleles than R alleles and did not increase the effective dominance of resistance.