Young novice drivers have a high risk of accidents due to their lack of routine automation. It is thus expected that they are more overloaded than experienced drivers in the same situation, with an increase of workload leading to performance impairments. This could be particularly observed in very complex situations. An experiment was carried out on a static driving simulator, with 16 traditionally trained drivers (aged 18 to 20) who had their driving license within the last two months and 16 drivers (aged 21) arriving at the end of the probationary period of three years. Three situations were presented in a counterbalanced order, i.e. a simple situation (a straight national road), a moderately complex situation (the same national road with bends), and a very complex situation (with twice the bends). Subjective workload was assessed by the NASA-TLX questionnaire after each situation while objective workload was physiologically recorded with the differential between the mean heart rate during the driving activity and before the activity (rest of 10 minutes). Driving performance was measured by the number of collisions with pedestrians who suddenly crossed the road in each situation. Results showed higher scores of physiological workload for novices in very complex situation, but no difference concerning subjective workload. Novices probably underestimated their physiological state because they undervalued the task difficulty and overvalued their abilities. Moreover, for all the drivers, the increase of subjective workload degraded their performance. Therefore, driving automation acquired with experience cannot always improve performance, especially faced with an unexpected hazard that requires flexibility in the behavior.