AbstractHost specificity determination of weed biocontrol agents has historically relied on evidence generated through quarantine trials in the region of introduction. These trials could give ‘false positive’ results due to a maximum type I error probability, and where possible, more research under field conditions should be conducted in the region of origin. The oligophagous, semiaquatic grasshopper, Cornops aquaticum Bruner (Orthoptera: Acrididae, Tetrataeniini), was released in South Africa for the biological control of Pontederia crassipes Pellegrini and Horn (Pontederiaceae). The aim of this study was to assess how the performance and field host range of C. aquaticum varies according to its stages of development, and how this contributes to the understanding of the relationship between the fundamental (laboratory‐based) and the ecological (field‐based) host range of this grasshopper, and its implications for water hyacinth biocontrol. We conducted post‐release laboratory no‐choice trials, confining early instars (instars 1 and 2), later instars (instars 3–6), and adult females and males in mesh cages, to determine insect performance on wetland plants growing in sympatry with P. crassipes. Also, gut analysis from field‐collected C. aquaticum was done to determine the ecological host range of this insect, identifying epidermal tissue of consumed plants. In no‐choice trials, survival rates of the later instars and adult C. aquaticum were similar on Pistia stratiotes L. (Araceae), Oxycaryum cubense (Poepp. & Kunth) Lye (Cyperaceae), and P. crassipes. However, under field conditions, P. crassipes and the congeneric Pontederia azurea Sw. were the only plant contents in the guts of early instars and the most abundant species in later instars and adults. The results support the hypothesis that C. aquaticum is an oligophagous insect on the genus Pontederia, and that different life stages should be considered when conducting host‐specificity trials in externally feeding mobile herbivore species. Diet composition of field‐collected insects thus could help detect false positives in laboratory trials, being an additional and realistic approach in understanding and predicting the selection processes of the insect in the new environment. Retrospective analysis of potential agents that were rejected due to lack of host‐specificity, using the methods from this study, could add a suite of additional agents to programs where invasive weeds remain unmanaged.