A sampling plan to estimate population levels of Liriomyza huidobrensis (Blanchard) larvae and total mining damage was developed for use in celery. Dispersion indices from Taylor's power law were calculated from 1992 data obtained from whole plant samples collected from commercial fields at two times during the cropping season and utilizing one of three different leafminer management strategies. Analysis of covariance indicated that variance–to–mean relationships were consistent among petiole position among plants, pest management strategies, and time of growing season for live L. huidobrensis larvae and total numbers of mines per petiole. In an attempt to develop a presence–absence sampling plan, all equation was derived to relate the proportion of petioles infested, P(l) , to the mean number of live larvae or the mean number of mines per petiole. However, predicted frequencies of infested petioles delivered from this equation were significantly greater than the frequencies of petioles infested with live larvae or infested with mines observed in validation studies. This lack of fit between predicted and expected frequencies precludes use of a presence–absence sampling plan to predict mean densities from the proportion of leaves infested. The results from Taylor's dispersion analysis were also used to evaluate a constant precision–level sampling plan. Validation tests of this sampling plan indicated that the formula used to ca1culate critical stoplines accurately predicted the observed cumulative totals of live larvae or total numbers of mines. Implementation of a sequential sampling plan for L. huidobrensis larvae with a 0.25 level of precision and with a possible maximum sample size of 100 petioles (to compensate for the prohibitively large sample sizes necessary at low densities) should greatly reduce the amount of labor necessary to achieve accurate estimates of population densities and to facilitate the use of action thresholds based upon these estimates.