1. The number of natural enemies that should be introduced to control a pest is a controversial subject in biocontrol. A previous semi-mechanistic model parameterised using a laboratory system consisting of two parasitoid wasps, Anisopteromalus calandrae and Heterospilus prosopidis, parasitising a pest beetle, Callosobruchus chinensis, indicated that the introduction of the non-native parasitoid H. prosopidis decreases the level of intraspecific interference between native A. calandrae females.The model also suggested that this decrease was the main factor destabilising the population dynamics of the host–parasitoid system, resulting in chaos. 2. To test this population-level decrease and host density independence in the interference of A. calandrae, we observed individual behaviours to quantify the level of intraspecific interference between two A. calandrae females in the presence or absence of H. prosopidis at two different host densities. 3. When H. prosopidis was present, the number of direct antagonistic interference events between A. calandrae females, sting duration, host feeding events (but not stinging events), and patch residence time were reduced. However, the presence of H. prosopidis decreased the patch residence time and the proportion of hosts parasitised by A. calandrae only when the host density was low. 4. The reduction in intraspecific interference between A. calandrae females by H. prosopidis and its host density independence support the population-level prediction, whereas the observed reduction in host-feeding behaviours in A. calandrae by H. prosopidis was not predicted. Overall pest control by the native parasitoid was unaffected by the non-native parasitoid as host density increased.