The aim of this work was to estimate sample sizes to assist the genetic improvement of the cashew tree (Anacardium occidentale L.). Stratified sampling, comprising five strata (S5, S4, S3, S2, and S1) of five cashew clones (BRS 274, BRS 275, BRS 226, BRS 189 and CCP 76), was effective for estimating the different sample sizes of the nut. Sample size for each clone depends on the weight-nut variance, the margin of error B permitted in the estimates and the desired precision of the results. The increases in sample size with clone variance, lowered the permitted margin of error B, and increased the desired precision of the results. These clones required different sample sizes for a morphological study of the nuts. Larger nuts require larger samples for the same margin of error B. For an error B of 0.2g, the sample size for clones S5, S4 and S3 were n5 = 84, n4 = 49 and n3 = 37 nuts. For clones BRS 274 (S5) and BRS 275 (S4), with better nut classification, the mean weights were respectively 16.79 and 12.78g. Clones BRS 189 (S2) and CCP 76 (S1), with smaller nuts, have a smaller variances, s22 = 0.7638 and s12 = 1.0712, where the mean weight was 8.29 and 7.81g respectively.