Drowsy driving poses a dangerous threat to road safety. This study aims to compare and evaluate the accuracy afforded by different modalities in assessing drowsiness, including: electroencephalogram (EEG), eye tracker, photoplethysmogram (PPG), and video recording. The work also investigates the impact of two parameters in the driving tasks on subjects’ level of alertness, presence of road-mark and frequency of lane-departure events. Ten healthy subjects without sleep deprivation participated in individual 1.5-hour experiments. A threshold of 80th percentile of reaction time was taken as ground truth to label data as Drowsy or Non-Drowsy. Using supervised learning random forest algorithm and stratified 10-fold cross validation, the results suggest that EEG features achieved highest classification accuracy: 0.957 and 0.864 for individual and combined sessions respectively, followed by eye tracker (0.821, 0.755 respectively). The highest accuracy of all modalities fell in the section that has the longest reaction time. These experiments additionally show that an absence of road-marks does indeed increase subjects’ reaction time, though they may not necessarily become drowsier. Further, low frequency of lane-departure events did make subjects drowsier as hypothesized. Since existing commercial products that claim to detect drowsiness are very expensive and target at vehicle transportation companies only, they are not available to daily private car users. As such, this type of study deserves attention so that drowsiness detection products could be made affordable and accessible to both professional drivers and daily private car users.