Three two-stage forest sampling procedures are examined in order (1) to quantify the variance penalty (if any) associated with second-stage sample restrictions; and (2) to quantify the variance gains (if any) associated with the incorporation of tree height data at the second stage. The three two-stage procedures share a common first stage, i.e., horizontal point sampling (HPS), which selects trees with probability proportional to tree basal area. The second stage selection approaches vary, as follows: 1. Trees are selected with replacement with probability proportional to tree height from a single list composed of all trees sampled across all first-stage sample points. 2. Trees are selected with replacement with probability proportional to tree height from separate lists compiled on each first-stage sample point. 3. Trees are selected randomly, with replacement, on each first-stage point. The results indicate that little or no reduction in variance accrues from incorporating height data into the two-stage sample design for the three mapped stands considered. HPS/simple random sampling yielded variances ranging from 7.1% larger to 9.2% smaller than the smallest HPS/list sampling variance, while maintaining an advantage in an even-aged stand. It is hypothesized that the HPS/list sampling procedures may prove more useful in mature, all-aged stands where height may account for a significant portion of volume or biomass variation. No loss of precision is noted when second-stage sampling is restricted by point on these three forest tracts. An investigation into the effects of first-and second-stage sample sizes provide a convincing argument to select only one tree per first-stage sample point when sampling is restricted by point. Considering both precision of estimation and field efficiency, these results suggest that the HPS/simple random sampling procedure is the most useful of the three tested. For. Sci. 40(2):247-266.