Sampling of Lepidoptera-infesting cotton bolls [Helicoverpa armigera (Hübner), Diparopsis watersi (Rothschild), Spodoptera littoralis (Boisduval), and Earias spp.] in Cameroon is based on examination of whole plants. To reduce the time taken by sampling procedures, we studied the suitability of a subsampling plan based on examination of plant terminals. The development of such a subsampling plan requires that the proportion of larvae located on plant terminals should remain constant. However, our study of the within-plant distribution of larvae revealed that their proportion on the five upper fruiting branches was too variable to allow the development of a sampling plan. Examination of the 10 upper branches led to less variable results, but the proportion was still significantly influenced by the pest species and time after crop emergence. We designed a conservative sampling plan in which the proportion was constant and equal to the lower bound of the 90% confidence interval for its predicted minimum (0.594). With this underestimation, the probability distribution of the number of larvae on the ten upper branches of a n plant sample was a compound of a negative binomial and a binomial distribution. These results enabled the design of a sampling plan that reduces sampling time by up to 60%, but the cost of this reduction is an increase in the risk in deciding on intervention when the mean infestation is lower than the critical density.