We recently developed a simple and inexpensive method that complements established bone histomorphometry procedures by enabling the two-dimensional imaging of cancellous bone to be viewed within its three-dimensional context with the marrow tissue in place and without detriment to the material for other histological purposes. The method, based on the preparation and superficial staining of slices 300 microm thick, enables "real" (i.e., unstained) trabecular termini to be separated from "artifactual" (i.e., stained) termini, providing a direct measure of cancellous connectivity in osteopenic bone. The technique was applied to osteopenic age-matched, white, postmenopausal women (31 with and 22 without vertebral compression fractures) with a similar bone status, as measured at the spine by absorptiometry and at the iliac crest by histology (see part I of this study). Despite the similarity in the mass of trabecular bone at either site, the results showed a significant difference (p < 0. 05) in the number of "real" trabecular termini between the groups, such that the fracture group had almost four times as many termini (mean +/- SE: 1.98 +/- 0.51/30 mm(2)) at the iliac crest as the nonfracture group (mean +/- SE: 0.53 +/- 0.31/30 mm(2)). Previous histomorphometry of the same material failed to detect a structural distinction between the two groups using established variables. It was concluded that a mass-independent trabecular discontinuity contributes to skeletal failure and that determination of the number of "real" disconnections (i.e., unstained termini) by the direct method proposed may provide a more sensitive discriminant of fracture than the present indirect procedures. A group of fracture and nonfracture men (see part I) suggested a similar distinction (fracture: 0.69 +/- 0.30/30 mm(2); nonfracture: 0.18 +/- 0.18/30 mm(2)), although the difference was not significant.