Abstract Variable retention has been advocated for maintaining biological diversity after regeneration harvest of forest stands. The Demonstration of Ecosystem Management Options (DEMO) study was established to test the effects of varying levels and patterns of residual trees on a range of forest taxa, ecosystem processes, and public perceptions. Understanding responses to experimental treatments depends on how well stand structural targets defining those treatments are achieved, but also on inadvertent changes in non-target structural attributes. DEMO treatments were specified by six levels and patterns of retained basal area: 100% retention (control), 75% aggregated retention (three 1-ha gaps cut within the treatment unit), 40% dispersed retention (regular distribution of residual trees), 40% aggregated retention (five uncut 1-ha aggregates), 15% dispersed retention (regular distribution of residual trees), and 15% aggregated retention (two uncut 1-ha aggregates). Treatments were applied randomly to 13-ha experimental units at each of six blocks in western Oregon and Washington. Treatment implementation produced the desired range of non-overlapping proportions of residual basal area. Other unspecified stand attributes such as tree density, stand density index, and canopy cover generally paralleled reductions in basal area, but quadratic mean diameter increased in the dispersed treatments due to selective retention of larger trees. Resulting stand structures were strongly dependent on initial conditions. Many differences in relative diameter distributions were observed before and after treatment, although they were changed less by aggregated retention. Strong differences in stand structure among blocks were highlighted by principal components analysis of diameter distributions, but experimental units converged on more similar structures at successively lower levels of retention. Indices of vertical complexity depicted by canopy area profiles showed strong responses to treatments, as did indices of horizontal variability in crown-area profile. Changes in tree species composition were small, although Douglas-fir ( Pseudotsuga menziesii) became less dominant at lower levels of retention. The combination of differing initial conditions and differing treatments created variation in residual stand structure that is independent of the categorical treatments and that may help explain residual variation in responses to these discrete treatments.