This study investigates the intricate relationship between task complexity and driving risk through a comprehensive four-phase on-road trial conducted in the UK. Employing Structural Equation Modelling (SEM), the research illuminates the factors influencing task complexity and its association with risk, treating both as latent concepts—unobservable variables in the study. The findings reveal a notable positive correlation between task complexity and risk, particularly concerning the headway indicator. In essence, the study demonstrates that an escalation in task complexity corresponds to an increased level of risk.Throughout the four SEM analyses performed across two waves of on-road trials, the time spent in each safety tolerance zone level for headway measurements emerges as a key indicator of the latent construct of risk in all phases. Notably, the variables constituting the latent concept of task complexity—those proven statistically significant—show slight variations across phases. Variables consistently significant across all phases include the number of right Lane Departure Warnings (LDWs) per 30 s and the day of the week.The models reveal the feasibility of quantifying the risk-task complexity relationship in real-world driving settings. This study provides insights to inform efforts to mitigate risk exposure through design and training interventions, targeting the most predictive factors linked to task complexity. Driver demographics did not emerge as statistically significant, emphasising the need for a holistic approach to improve road safety.