A new taper model was developed based on the switching taper model of Valentine and Gregoire; the most substantial changes were reformulation to incorporate estimated join points and modification of a switching function. Random-effects parameters were included that account for within-tree correlations and allow for customized calibration to each individual tree. The new model was applied to 19 species groups from data collected on 2,448 trees across the northeastern United States. A comparison of fit statistics showed considerable improvement for the new model. The largest absolute residuals occurred near the tree base, whereas the residuals standardized in relation to tree diameter were generally constant along the length of the stem. There was substantial variability in the rate of taper at the base of the tree, particularly for trees with relatively large dbh. Similar trends were found in independent validation data. The use of random effects allows uncertainty in computed tree volumes to compare favorably with another regional taper equation that uses upper-stem diameter information for local calibration. However, comparisons with results for a locally developed model showed that the local model provided more precise estimates of volume.