Although the traditional ''lie detector'' test is used frequently in forensic contexts, it has (like most test of deception) some limitations. The concealed knowledge test (CKT) focuses on participants' recognition of privileged knowledge rather than lying per-se and has been studied extensively using a variety of measures. A ''guilty'' suspect's interaction with and memory of crimescene items may vary. Furthermore, memory for crimescene items may diminish over time. The interaction of encoding quality and test delay on CKT efficiency has been previously implied, but not yet demonstrated. We used a response-time based CKT to detect concealed knowledge from shallow and deep study procedures after 10-min, 24-h, and 1-week delays. Results show that more elaborately encoded information afforded higher detection accuracy than poorly encoded items. Although classification accuracy following deep study was unaffected by delay, detection of poorly elaborated information was initially high, but compromised after 1 week. Thus, choosing optimal test items requires considering both test delay and initial encoding level.