In biological control projects, establishment of released natural enemies is a key step and must be efficiently detected. We studied the relative efficacy of larval versus adult sampling to detect establishment on spotted knapweed, Centaurea maculosa Lamarck, of the two root feeding insects Agapeta zoegana L. and Cyphocleonus achates (Fahraeus). Larval sampling was based on excavation and dissection of plant roots. Adult sampling consisted of either sighting adults along transects at release sites or collection of adults by sweep netting. Recovery rates for A. zoegana were higher through adult visual sampling (54.8%) than through larval sampling via root dissection (43.0%). Adult visual sampling required less time (30 min/site) than did root dissection (130 min/site). Sweep net sampling, although having the lowest detection rate (38.1%), required even less time (10 min/site) and was the most effective method per unit time. In contrast, for C. achates, larval sampling was the most effective method, with a recovery rate of 35.6%, compared with 8.9% for adult visual sampling. Sweep netting was more effective than visual sampling, with a detection rate of 18.1%. Adult visual sampling required less time (44 min/site) than did root dissection (130 min/site). Sweep net sampling, although having the lowest detection rate (18.1%), required even less time (10 min/site) and was the most effective method per unit time. However, other factors such as weather, travel time, and training levels needed for sampling make root sampling a more effective method, in a larger sense, for both of these insects.