The cognitive interview (CI) is a set of forensic evaluation procedures that has proven its efficiency. Nevertheless, little investigation has been made into its usage on witnesses with the intention of lying. With the aim of proving whether the efficiency of the CI is maintained where the witness is lying we carried out two experiments and we included instructions to tell the truth or to lie as independent variables. We used a 2 Type of interview (SI/CI))2 Statement Condition (True/ False))2 Interviewee Gender (Male/Female) design with 18 interviewers in a sample group of 240 gender-matched university students being interviewed and the classic measures from the CI as dependent variables: the total number of details, details of actions, details about people and details concerning objects. In a first experiment, participants who lied exculpated the protagonist of a film fragment from rape, whereas in a second, those who lied accused an innocent protagonist. The results, in accordance with previous studies, showed a globally better efficiency for CI. Nevertheless, analysing the material based on type of details, this efficiency is only found in details concerning actions and objects. We discuss the forensic importance of these findings.