An initial experiment was carried out which aimed to evaluate a particular function delegation mode on the test track. A car was equipped with an automatic device in order to replace the driver's lateral control of the trajectory (driving in the lane centre). The longitudinal control (speed) remained under manual control. The results showed difficulties in returning to manual control outside the domain of validity of the automatic device, with a suspicion of complacency. However, longer experience with the device would have been necessary to reach a higher level of trust and performance in the interaction between the driver and the device. In addition, negative interference between the bend-taking styles of the device and the driver were identified. A second experiment was carried out, again on the test track, in order to evaluate two types of mutual control modes. This took the form of criticising the driver when the car deviated more than 60 cm from the lane centre. With the warning mode, the driver heard a lateral noise on the side of the deviation and felt a non-specific steering wheel oscillation. With the action suggestion mode, the driver felt a steering wheel oscillation; this made the driver turn it toward the opposite side of the deviation. The two modes have revealed themselves to be sometimes efficient in critical situations where a trajectory deviation beyond 60 cm was provoked by a visual occlusion. Individual differences in effects were larger with the action suggestion mode than with the warning mode. No adverse cooperation effects were noticed.