Coaching, not instructing: teaching the skills beyond car control

Traditional lessons are built around two goals: work the car, and obey the rules. Clutch, gears, mirrors, roundabouts, road signs: the visible, testable mechanics of driving. And learners do need these. But when researchers ask why young drivers actually crash, car control is only part of the answer. Studies of crash reports point to inexperience and failing to spot hazards, and the GDE matrix's authors add decisions and motives: the decision to follow too close, to speed up for a friend, to drive angry or exhausted, to believe the rules are for other people. Those decisions sit at a different level, and standard instruction barely touches it.

Decades of evaluations found that traditional driver education, on its own, did little to reduce crash rates. The 2024 NHTSA systematic review of promising practices notes the gap described above: it reports a survey (Bartl et al., 2005) which found that most European countries' standards for driving instructors focused almost exclusively on the lower stages of the GDE matrix, vehicle control and driving in traffic.

The GDE matrix, which came out of the EU GADGET research project, describes competent driving as four levels in which the higher levels guide the ones below. What you do at a junction (level 2) follows from the choices you made about the trip (level 3) and from who you are and how you live (level 4). It is not a simple one-way chain, though: the levels are constantly checked against feedback from the driving itself, and changes lower down, such as new or imagined skills, affect the whole system (Hatakka et al., 2002; Peräaho et al., 2003).

The matrix also has a column that changes everything: self-assessment. At every level, a driver should be able to judge their own strengths, weaknesses and risky tendencies. This is the thread that ties the GDE approach to calibration, and it is the column that can only be developed by asking the learner to reflect, never by telling them the answer.

Coaching means shifting from delivering answers to drawing them out: asking questions, prompting reflection, and handing the learner responsibility for spotting and solving their own faults. In one randomised trial (Passmore & Rehman, 2012), 208 serving members of the British Armed Forces who were learning to drive a lorry were taught either by instructors trained in coaching, who blended coaching with instruction, or by instructors using instruction alone. The coached group spent less time in training and were more likely to pass first time. It is a single study, its learners were lorry trainees rather than car learners, and it measured training and test results, not crashes. A driver who works out why for themselves owns the skill more deeply than one who was simply told.

The bigger prize is at the higher levels. The EU MERIT and HERMES projects call coaching an optimal method for addressing attitudes towards risk (level 4), and HERMES argues that for higher-level issues coaching techniques are ahead of teaching methods. That is a position, not a tested result, and HERMES also says coaching should supplement teaching, not replace it. The reasoning is that you cannot lecture someone into better self-awareness or safer motives. They have to arrive at it themselves, guided by good questions.

When a learner makes a mistake, the instructor's reflex is to supply the fix: "you were too fast there." Coaching interrupts that reflex and substitutes a question: "how did that corner feel? What would you change next time?" The learner does the diagnosis. It's slower in the moment and far faster over the course, because self-generated insight sticks while corrections evaporate.

A simple structure helps: ask what they noticed, what they'd do differently, and (the higher-order move) why it matters for the kind of driver they want to be. That last question quietly reaches up into levels 3 and 4, where instruction never goes. The instructor's job becomes managing the conversation and the risk, not narrating every input.

1 · Abandoning instruction entirely. A brand-new learner grinding the gears doesn't need a reflective question. They need to be shown. Coaching is layered onto secure basics, not used instead of teaching them. Match the method to the level.

2 · Asking fake questions. "Was that a bit fast?" isn't a coaching question. It's a correction wearing a question mark. Real coaching questions are open and hand the thinking to the learner. If you already know the answer you want, you're still telling.

3 · Coaching through danger. When something is unsafe, act and instruct clearly and immediately. Save the reflective conversation for afterwards. Coaching manages learning; it never overrides safety.