Mapped-stand data which includes tree location and mensuration data may be used to simulate the height characteristics of a forest canopy in a computer. The dimensions of the rectangular, fixed-area ground plot which is used to collect the tree location and mensuration data can profoundly affect the accuracy of the resultant computer reconstruction of forest canopy heights. Thin fixed-area plots result in forest canopy height constructs (i.e., canopy height models, or CHMs) which are shorter and less dense than the actual forest stand. Ninety-nine 0.405 ha forest stands in the southeastern and south central United States were analyzed. For each stand, a top-of-canopy height model was generated for the entire stand on a 0.25 m x 0.25 m grid. The mean canopy height and the canopy density of the simulated stand was calculated. A Monte Carlo simulator established randomly located sample plots of varying plot width (1 m to 20 m) at three different sampling intensities (1%, 5%, and 10% of stand area) in the mapped stand. The trees selected on any particular plot were used to generate a CHM, and the sample mean canopy height and canopy density were compared with the corresponding stand values. Samples collected using 1 m wide plots resulted in height models which underestimated average canopy height and canopy closure with deviations greater than 5% of the stand value, 95-100% of the time. The underestimation results from the construction of top-of-canopy architecture based on measurements of crowns belonging only to trees whose boles are within the plot. This construction technique omits crowns extending over the plot belonging to tree boles located off the plot. Thinner plots tend to miss more of the trees which support the canopy over the plot. For the 99 stands considered in this study, ground plot widths on the order of 6 to 8 m were required in order to acquire ground data which could be used to accurately model forest canopy height characteristics. The minimum ground plot width required to produce an accurate height model decreased with increasing stem density; a plot width of 4 m, for instance, may be used at stem densities of ~2000 stems ha-1. For. Sci. 44(3):438-444.