Plants of each accession were cultivated in a field setting in Ames, IA or at Wageningen, Netherlands. All xylem material came from 2-year-old twigs growing on sapling-size or larger trees. The twigs were of comparable size. The material from The Netherlands was collected in early April and that from Ames, in mid-June. Lengths of twigs adjacent to twig crotches were killed and fixed in formalin-alcohol-acetic acid solution (FAA) (Sass, 1958). Isopropyl alcohol was used in The Netherlands and ethyl alcohol in Iowa. All twig material was stored in FAA for up to 2 years before cross-sections were made. For sectioning, the fixed twigs were wrapped tightly in paper towelling and cut 15 to 20 um thick with a sliding microtome. The cross-sections were stored in distilled water at 4C before anatomical observations were made. Three of the Ames selections were also embedded in paraffin, cut 10 um thick on a rotary microtome and stained with saffranin-fast green (Sass, 1958). The latter was done to check the accuracy of vessel determination in the thicker, unstained sections. Observations were made on images of twig cross-sections projected on white cardboard using a projection mirror mounted on a compound microscope. With the microscope placed at stand-up bench height, the image was projected to sit-down table height. An enlargement of x 248 was obtained thereby. Determination of vessels was aided by the appearance of bordered pits common to these cells. Vessel lumen diameter was measured in the projected image, then converted to micrometers on the sample. The diameter of a vessel group was estimated as the product of the average vessel diameter and the average number of vessels per group. The data represent the vessel characteristics of a constant area of xylem grown the year prior to collection (second-year growth). All earlywood and most, if not all, late wood were included in this area. For each cross-section observed, data were recorded for four different areas. Two cross-sections were checked from each of two branches collected from each elm selection. Only one tree for each selection was used. The data from the sixteen areas for each elm selection were averaged. Although variability within a selection was present, it did not affect statistically the final comparisons among the different selections.