The sampling dynamics of the pink bollworm (Pectinophora gossypiella (Saunders)) were studied during the cotton growing seasons of 1976, 1977, and 1978. Different sampling methods were employed each year to compare sampling intensity and to compare whole plant samples with susceptible boll samples. Analysis of variance data indicated that significant differences in mean boll infestation levels occurred between small areas of a field (ca. 1 acre), but that these differences were masked when larger units (ca. 10 acres) of a field were used as the sample area. From these data it was found that samples yielded the most accurate pink bollworm infestation estimate when taken from as large an area of the field as possible. Data indicated that accuracy improved only 2 to 3 percent when sample size increased from 60 bolls to 100 bolls. Bolls taken from whole plant samples were as accurate as susceptible boll samples if there was an equal number of bolls in each sample. Sample accuracy was nearly stable for fields 20 to 40 acres in size. A sequential sampling plan for pink bollworm was developed using the binomial distribution. Decision levels were established using the 6 and 12 percent boll infestation levels with α and β levels of 0.2. Approximately 50 percent of the sampling time can be saved over conventional 100 boll samples with very little loss of accuracy for a spray-no spray decision based upon an economic threshold value of 15 percent pink bollworm infestation.