AbstractNon-target effects of approved biological control agents have raised questions about the safety of biological control of weeds and resulted in an increased emphasis on monitoring and reporting of non-target effects as part of post-release assessments. This is particularly important in the case of the root-mining weevil Mogulones cruciger (Herbst), which was approved in Canada to control houndstongue, Cynoglossum officinale L., but not in the United States because of concerns over its environmental safety. To address these concerns and the potential for non-target effects, we monitored co-occurring confamilial Boraginaceae species at six M. cruciger release sites in Alberta and British Columbia over two years. All four co-occurring species were attacked by the weevil to varying degrees although attack was inconsistent between years and sites, and non-targets were mostly attacked to a lesser degree than houndstongue. There was a positive relationship between the probability of non-target attack and houndstongue attack rate by M. cruciger indicating potential spillovers and early evidence suggests non-target attack may be transitory. The comparison between the pre- and post-release evaluations and preliminary plant volatile, electroantennogram, and host-choice behaviour data suggest that chemical ecology may provide an important tool in understanding an insect's host-choice selection in pre-release host-specificity assessments.