This thesis examines factors limiting understory herb presence and flowering in young second-growth Douglas-fir (Pseudotsuga menziesii) forests on the west side of the Cascade Mountains, Oregon, USA. I studied the belowground effects of canopy trees on understory herbs and shrubs in old-growth forests using trenched plots from which tree roots were excluded. Effects of tree density and stand age were tested by comparing the understory community composition of old-growth stands and pre-commercially thinned and unthinned young second-growth stands. I also examined the effect of conifer basal area on understory herb presence and flowering within one young second-growth watershed. In young stands, I focused on three groups of understory herb species: disturbance-responsive (release), forest generalist and old-growth associated. The effects of root trenching on vegetation and soil moisture were tested in closed-canopy and gap locations in two old-growth Douglas-fir (Pseudotsuga menziesii) forests. Ten years after installation, trenched plots averaged 92% total understory cover while untrenched plots averaged 47% cover. Trenched plots under closed canopies were moister than control plots throughout the growing season; the trenching effect on soil moisture became apparent in the generally wetter gaps only at the end of the growing season. Vegetation responses to trenching were concomitantly larger under closed canopies than in gaps. Stands that had been pre-commercially thinned 20 years earlier exhibited understory composition more similar to old growth than did unthinned stands. Thinned stands exhibited higher frequencies, abundances and density of flowering of old-growth associated herbs than did unthinned stands, but lower than did old-growth stands. Forest generalist and release species showed mixed responses to thinning. I used both general linear models and classification and regression tree models to explore the association of herb species presence and flowering with conifer basal area and abiotic variables. Both modeling approaches yielded similar biological insights. Flowering was more sensitive than presence to current stand basal area. Flowering of old-growth associated and release species was negatively correlated with conifer basal area. Linear models allowed clearer hypothesis tests, while tree-based models had greater explanatory power and provided information about interactions between variables. / Graduation date: 2004