SUMMARYA satisfactory method of collecting samples ofThrips nigropilosuswas developed in which only the mature leaves were collected, and a number of plants were sampled across several rows of the plot. This give a reproducible and representative estimate of the population within the plot.Despite the development of this technique, very great differences remained between counts on plots of a Latin square experimental design. The use of covariance analysis considerably reduced the residual variance when populations were increasing arithmetically or at only slightly increased rates. When the increase became more rapid, the value of covariance diminished rapidly while simultaneously positional effects became more significant, suggesting that restraining influences were operating unequally in different parts of the experimental area. It is considered that this phenomenon is widespread in populations ofT. nigropilosusand that, as expanding populations are the most likely to be subject to experiment, a more complex design, probably of the incomplete block type, should be used in future.