This article lies within the field of applied cognitive linguistics (ACL) and presents empirical work that addresses overlooked effects of assessment typology in second language (L2) learning. It examines whether pairing a cognitive instructional approach with matching assessment design results in greater learning outcomes over the more pervasive notional-functional approaches. The last two decades have witnessed a proliferation of empirical research measuring the effectiveness of ACL-based teaching approaches, yet studies have only been partly fruitful in eliciting data that truly favors ACL. We argue that this is largely due to assessment design, which typically measures performance via correct vs. incorrect tasks. To overcome this caveat, two studies addressing the complex Spanish psych-verb construction (e.g. gustar ‘to like’) were conducted following a pretest/posttest/delayed-posttest design for three empirical conditions (control, cognitive, and traditional): a pilot study (n = 59) and a larger replication (n = 160). Data collection entailed ACL-based assessment for interpretation and production tasks. Results showed that after instruction, the cognitive group significantly outperformed the traditional counterpart in both tasks. These findings lend support to the effectiveness of pairing cognitive instruction and cognitive assessment for difficult grammatical constructions.