Recently ecologists have suggested that physical ecosystem engineers may be effective conservation targets because of their potential to foster community development. However, because ecosystem engineers are often long-lived, slow growing species, few studies have explored whether interactions between engineers and associated species result in predictable patterns through time. Kelp forests offer a unique opportunity to explore how ecosystem engineers influence community development. Like trees in terrestrial forests, kelp negatively affects some understory species and positively effects others. Unlike terrestrial forests, frequent disturbance and fast growth rates cause the abundance of giant kelp to vary dramatically on short time scales. First, I examined the relationship between temporal variability in kelp abundance and understory community composition using observational data collected on natural reefs in Santa Barbara, CA. I found that sites where kelp abundance was highly variable between years were sites where understory species assemblages were highly variable, and vice versa. In fact, standard deviation in kelp abundance predicted over 65% of the variation in community structure. Second, I used data from a system of artificial reefs in San Diego, CA to test the effects of kelp on community development. I found abnormally high kelp abundances on artificial reefs (~2x natural reef average), which led to high abundances of species positively affected by kelp (invertebrates) and the exclusion of negatively affected species (algae). These data suggest that using ecosystem engineers in restoration projects can be effective, but that abnormally high abundances of engineers can exclude species that are normally present on natural reefs.