Practising with a parent: what 22 months of in-car cameras showed

Between 2011 and 2014, researchers at the Virginia Tech Transportation Institute instrumented the cars of 82 teenagers in Virginia. Each car carried four cameras plus accelerometers and GPS. Every participant was recruited within three weeks of getting their learner permit, at an average age of 15.6, so the recording starts before the first practice drive rather than halfway through. The kit stayed in the car through the whole permit period, which averaged 10.35 months, and then through the first twelve months of independent driving after licensing. That is roughly 22 months of continuous recording per family.

The volume is what makes it usable. 18,686 trips during the permit phase. About 110,000 miles of supervised driving and about 380,000 miles of independent driving, 490,000 miles in total. Nine crashes were recorded during the permit phase, and 69 across the permit and independent phases together. Note those crash counts, because they come back later when we get to the honest limits.

The footage is old. The analysis is not. It was published in April 2025 as BTSCRP Research Report 13, "Outcomes of Variability in Teen Driving Experience and Exposure", by Sheila Klauer, Lesheng Hua, Shu Han, Feng Guo, Johnathon Ehsani, Michelle Duren, Loren Staplin and Tia Mastromatto, for the National Academies. If you saw a headline last year about cameras proving that practice works, this is the study behind it, and the cameras themselves stopped rolling over a decade ago.

The headline number, and exactly what it is

Split the 82 into two halves at the median of their normalised monthly practice hours, then compare their rate of crashes and near-crashes per 1,000 hours once they were driving alone, across the first twelve months of it. The less-practised half recorded 17. The more-practised half recorded 12. Virginia Tech's release, dated 27 May 2025, rounds that gap into the sentence that travelled: teen drivers who practised more had "30 percent fewer crash or near-crash incidents". The arithmetic is a 29.4% relative reduction between two observed rates. It is not a modelled effect size with a confidence interval, and you should not treat it as one. The gap is also not in the practice itself. During the permit phase the two groups ran almost identical rates, 0.0057 against 0.0056 events per hour. The difference only appears after the L plates come off.

Hold that quote lightly on the "variety of environments" half. The amount of practice and the variety of practice were measured with different outcomes and came out with very different strength, and Section 5 pulls them apart properly. The report itself is more cautious than the press release, which is normal and is not a scandal.

Across the whole permit period, averaging most of a year, the less-practised half averaged 17 hours of supervised practice each, 698.23 hours across that whole group. The more-practised half averaged 52.7 hours each, 2,160.54 hours across the group. Normalised over nine months, the average across the full sample came to 29.6 hours of permit driving. So the gap that produced the 12 versus 17 difference is roughly the difference between two hours a month and five hours a month.

Virginia requires 45 hours of supervised practice, 15 of them after sunset, for anyone licensing under 18. Only about half the sample got to 45. The report says it plainly: "at least half of all teens may not be getting the full number of hours of learner's permit driving." A legal minimum on paper turned out to be an aspiration in practice, in a state that had one.

Ireland does not have one. There is no mandated number of supervised practice hours here at all. The only compulsory professional element is Essential Driver Training: 12 lessons of one hour each, tied to 12 high-risk topics, delivered by an approved instructor. Everything beyond that is voluntary practice with a sponsor. What the RSA does provide is a recommendation, and CIECA's account of the programme states it cleanly: the RSA recommends "that the learner driver practises the skills learned in each compulsory EDT lesson with their Sponsor for at least three hours before their next lesson."

Do the arithmetic on that and it is our own sum, not an RSA figure: three hours after each of twelve lessons comes to about 36 hours. That lands between the study's less-practised group at 17 hours and its more-practised group at 52.7. It is a respectable target, it is a recommendation rather than a requirement, and in our experience almost nobody hits it. The bit that stings is that 36 hours spread across a six-month permit is about an hour and a half a week.

If you want the detail on what the twelve lessons actually cover and in what order, our EDT lessons explained page walks through each one, and how many driving lessons you actually need in Ireland deals with the follow-up question everybody asks, which is whether twelve is enough. Short version on that: for most people it is not, and the reason is in this article.

Ehsani and colleagues published on this cohort in JAMA Pediatrics in 2020. Their descriptive numbers are worth seeing next to the report's: a mean of 48.2 hours of practice, with a standard deviation of 33.3, over 1,259.2 miles and 88.9 separate driving days, across a mean permit period of 10.35 months. Note that the two papers do not agree on the hours. The BTSCRP report's own group totals work out at a mean of about 35 hours for the same 82 teenagers, so the two teams counted practice time differently, and neither figure should be quoted as the definitive one. What both agree on is the standard deviation being the interesting part. The spread between families was enormous, far bigger than either average tells you.

They then built a measure of consistency: the number of days the learner actually drove divided by the number of days they held the permit. Higher consistency was associated with lower crash risk in the first year of independent driving, with a hazard ratio of 0.61 (95% CI 0.38 to 0.99) for each 0.23-unit rise in that score. Higher sensation seeking pushed the other way, at a hazard ratio of 1.67 (95% CI 1.08 to 2.57) per 0.75-unit rise. For scale on the outcome, 28 of the participants recorded 49 property-damage or police-reportable crashes in that first solo year.

Be honest about that confidence interval. The upper bound is 0.99. It clears 1 by a whisker in a sample of 82. This is a signal worth acting on because it is cheap to act on and it points the same way as everything else in this article, not because it is bulletproof.

Translated into a diary: forty minutes most weeks beats a six-hour blitz the fortnight before the test. Two short drives a week, even twenty minutes each, keeps the skill warm and keeps the habit alive. And there is a second-order reason to care about consistency that the statistics do not capture. A learner who dreads getting in the car with you stops getting in the car with you, and then your hours collapse regardless of what you intended.

Practice happened on roads the learner already knew. 89.77% of trips in the more-practised group and 86.04% in the less-practised group were on familiar routes. Call it nine trips in ten. About 40% of practice hours fell at weekends. Roughly 18% of practice hours were on controlled-access or high-speed roads, which worked out at an average of 7.9 hours each for the more-practised group against 2.5 hours each for the less-practised.

Night practice was the clearest failure. Only about 10% of all practice happened at night. In hours, that was an average of 1.9 hours each for the less-practised group and 7.4 hours each for the more-practised group, against the 15 hours after sunset Virginia requires. Both groups missed it, one of them by a mile.

The report also found that hard braking was more frequent at night and on unfamiliar routes. That is exactly what you would expect, and it is exactly the point: the difficulty is real, so somebody has to be there the first time the learner meets it. The alternative is that they meet it alone, on their own, at twenty to midnight, six weeks after passing, with a car full of friends.

The comparison figures for that are American and are about a different licensing system, so take them as direction rather than as Irish numbers. Among US drivers aged 16 to 19, the fatal crash rate at night runs at about three times the adult rate per mile driven, and in 2022, 20% of US teenage motor vehicle crash deaths happened between 9pm and midnight with a further 16% between 6pm and 9pm. Closer to home, An Garda Síochána counted 190 road deaths in 179 fatal collisions in 2025, five of them in public places such as car parks. On the RSA's public-roads count the figure was 185 deaths in 174 fatal collisions, up 8% on 171 deaths in 157 collisions in 2024, and the highest total since 2014. Drivers accounted for 76 of those deaths, 41% of the total, and people aged 16 to 25 accounted for 38, still the highest-risk age group.

The evidence we lean on is the Cochrane review of post-licence driver education, CD003734: 24 trials, close to 300,000 drivers, a pooled crash relative risk of 0.98 with a 95% confidence interval of 0.96 to 1.01 across the 15 trials that recorded crashes. For practical purposes, no effect. That review is about advanced and remedial courses taken after licensing. It did not test EDT, it did not test pre-test tuition, and nothing in it says an instructor is useless.

The BTS-23 finding is about a different thing entirely: hours in the car with a sponsor between lessons, before the test. So the honest reconciliation is that these are two separate interventions at two separate stages, and the crash-relevant evidence for the practice is currently in better shape than the crash-relevant evidence for the tuition. That is an uncomfortable sentence for a driving school to publish, which is precisely why we are publishing it.

Here is what we think the division of labour actually is, stated as our reasoning rather than as a research finding. Professional instruction supplies the method: the correct routine at a junction, the correct observation sequence, the correct way to handle a roundabout, and a competent adult noticing the fault you cannot see yourself. Supervised practice supplies the mileage that turns a method into a habit. Neither substitutes for the other. Twelve hours of excellent instruction with no practice between lessons produces a learner who knows what to do and cannot yet do it. Fifty hours of practice with no instruction produces a learner who has thoroughly embedded whatever their dad happens to do.

If you want the longer version of our position on training evidence, it is set out in does driver training actually work, and the hazard-perception piece at what hazard perception training does and does not do is where the Cochrane figure sits on this site with its full citation.

Start with the European framing. The Commission's ERSO synthesis on novice drivers notes that in formal training schemes "young drivers often only have about 25 to 40 hours of driving experience when they are licensed for solo driving", sets that against evidence from Gregersen and Sagberg that risk would be greatly reduced with far more pre-licence experience, and lists the remedy bluntly among its countermeasures: "Introduce high levels of pre-licensing accompanied practice. This is potentially one of the most effective countermeasures." The same document records that drivers aged 16 to 24 carry "2 to 3 times higher risks than those of more experienced drivers", and that their crashes typically happen at night, are single-vehicle, and involve loss of control and high speed.

Sweden ran the natural experiment. In September 1993 it lowered the minimum age for supervised practice from 17.5 to 16 while leaving the licensing age at 18, effectively doubling the available practice window. Novice accident risk fell by about 15% overall, and by about 40% among those who used the new lower age, with the early starters getting roughly 2.5 to 3 times as much supervised practice, averaging around 118 hours. Treat the 40% carefully: the people who chose to start at 16 chose it, so that subgroup selects for exactly the sort of family that would have practised more anyway. The 15% figure across the whole novice population is the more conservative one, and it is the one worth quoting.

CIECA's synthesis of the accompanied-driving evidence puts a distance on it rather than an hour count, reporting that experience in Sweden and Norway suggests high levels of pre-licence accompanied practice can significantly reduce accident risk in the initial post-licence phase, and that "a considerable 5000-7000kms of accompanied driving experience are necessary to achieve this aim." It also makes a point that answers the obvious parental worry: the accompanied phase itself produces "only marginally more accidents when compared to practical lessons with a driving school." Practising with your parent is not the dangerous bit.

Where more hours stopped helping

Now the counterweight, which is the most important paragraph in this section. Kettlewell and Siminski at UTS analysed all New South Wales licensing and crash records across two separate reforms. Raising the mandated practice requirement from 0 to 50 hours cut the probability of a serious accident in the first year of solo driving from 6.9% to 5.4%, a relative reduction of about 21%. Raising it from 50 to 120 hours produced no measurable safety benefit at all. And even the 50-hour gain did not persist beyond the first twelve months of solo driving. Victoria, meanwhile, requires 120 logged hours before a learner under 21 can sit the drive test.

This year's systematic review lands in the same place. Koppel and colleagues, writing in the Journal of Safety Research in June 2026, reviewed 18 peer-reviewed studies published between 2003 and 2024. Longer learner-permit periods and structured training came out protective, and greater supervised practice correlated with safer outcomes, but "evidence isolating specific hour or distance thresholds was mixed."

Two things follow, and neither is the lazy conclusion. First, do not say there is no magic number of hours. Mixed evidence on thresholds is not proof that no threshold exists, and the BTS-23 authors themselves conclude that the hours states recommend "should probably be no less than the 45 hours required by the state of Virginia", while calling for research on what the optimal number actually is. Second, the direction of travel is not in doubt: more practice is better than less, the gains look front-loaded in the first 50 hours or so, and the returns above that are unproven. Note also that copying a scheme does not copy its results. ERSO records that a French accompanied-driving scheme comparable to the Scandinavian ones "failed to be effective". How the practice is done matters, not just how long it lasts.

An Garda Síochána sets out the accompanying driver requirement precisely: the accompanying driver must hold a full licence in the same vehicle category, and must have held it continuously for at least two years. They sit in the front passenger seat. A learner permit holder cannot accompany another learner, and neither can somebody eighteen months into their own full licence, which rules out a lot of older siblings. L plates go front and rear. In practice there is one more constraint that nothing official will remind you about: the sponsor normally has to be a named driver on the policy for the car being used, so check the insurance before you plan anything.

Then the one that matters for this article. Learner permit holders cannot drive on motorways under any circumstances, accompanied or not. In our patch that removes the M50, the M7, the M4 and the M3 from every practice plan. It also means that the roughly 18% of practice hours the camera study logged on controlled-access and high-speed roads is not something an Irish learner can copy, and that motorway competence is necessarily a job for after the test, during the N-plate period. Our Irish motorway driving guide is written for exactly that moment.

You can get most of the benefit legally. National and regional roads, and dual carriageways that are not motorways, give you higher speeds, merges, slip roads and overtaking without breaking the rule. Four routes in our patch, with the trap on each one named:

• N51 west out of Navan towards Athboy, about 15 km of national secondary road. The first few kilometres are 100 km/h dual carriageway, and roughly 3 km out you drive straight through M3 junction 9 at Boyerstown. Keep following the green signs for Athboy and Delvin. Never take a slip road with a blue sign on it.
• N81 out of Tallaght towards Blessington, national secondary the whole way. Join it on the Tallaght side and drive away from the city, because the N81 begins at the M50 at junction 11. Build up to this one rather than starting on it: the collision record on the Wicklow stretch is poor.
• N2 north of the Rath roundabout, above Ashbourne. The M2 ends at that roundabout, about 1.5 km north of the town, and the N2 onwards towards Kilmoon Cross and Slane is an ordinary national road. Get there through the town on the R135, not on the dual carriageway that bypasses it, because the bypass is the M2.
• R148 through Leixlip and Maynooth to Kilcock, the old N4, regional road throughout. Note that the R148 does not reach Lucan: leave Lucan on the R835 Leixlip Road and it becomes the R148 at Cooldrinagh. Do not let a sat nav plan Lucan to Maynooth for you, because it will send you onto the M4 at junction 5.

All four are legal for a learner and all four are completely different from an estate loop. The wider point is to read the signs at the junction rather than the map, because around here the motorway entrances sit on the edge of town and a sat nav will find one within a minute of setting off. That does not mean those towns are cut off. Navan to Kells: the M3 is the fast route and it is out of bounds, but the old N3 is now the R147 and runs the whole 16 km, signed Kells Road on the way out of Navan. Ashbourne towards Dublin: the M2 has an entrance at the Rath roundabout and another signed Ashbourne South, so watch for blue at both, but the old N2 is now the R135 and it will take you to Finglas by way of The Ward. Blue sign with a white M means turn back. Green or white signs with an N or R number are yours.

The full detail on who qualifies to sit beside a learner, and the awkward cases, is on our dedicated page: qualified accompanying driver rules in Ireland.

The RSA recommendation of three hours between lessons is easiest to hit as three separate sessions of about an hour rather than one long one, and the consistency finding says the split version is the better shape anyway. Give each session a different job, and the variety looks after itself instead of needing to be remembered.

Across the whole permit period, keep a running list of conditions rather than only a running total of hours. The camera study's own weakness was that the hours went in and the variety did not, and the report even suggests electronic logging apps as a way to make compliance visible. A pen in the back of the logbook does the same job.

One more piece of timing. Because Irish test waiting times mean the gap between finishing EDT and actually sitting the test is often months, that stretch is dead time unless practice continues through it. Check the current position on our driving test waiting times page and plan practice through the wait rather than stopping when the twelfth lesson is signed off. If you want the lesson-by-lesson structure to hang sessions off, the full course is laid out at learner lessons, stage by stage.

What not to do on a sponsor drive

Bad practice is not neutral. It is rehearsal, and whatever gets rehearsed is what gets kept. Five specifics, in the order they cause us the most work.

Do not teach a different method from the instructor. This is the big one. The EDT logbook exists partly so the sponsor can read what the ADI wrote after the last lesson and practise that. If the instructor teaches a mirror sequence and you teach the one you learned in 1994, the learner now has two systems and will fall between them under test pressure. Only the ADI can sign and stamp the EDT lessons; the sponsor's job is to reinforce, not to redesign.

Do not let something be practised wrong for a fortnight. A manoeuvre repeated incorrectly for six sessions costs more instructor time to unpick than the six sessions gained. If you are not sure whether what they are doing is right, that is a question for the next lesson, and the honest answer is to leave that manoeuvre alone until you have asked.

Do not run the same loop for months. The camera data pins this one down: it is the most common shape of practice and it is the one that stops paying earliest. If you could drive the route with your eyes shut, so can they.

Do not let the drive become an argument. A learner who dreads practice practises less, and consistency was the protective factor. Correcting one thing per drive, calmly, at the end, beats a running commentary. If it goes badly, end the session early and go home rather than pushing through to prove a point. The Ehsani finding that errors during practice may be learning events is your permission slip to let small mistakes happen without a lecture.

Do not skip night practice because night driving is more dangerous. This is the argument that sounds most responsible and is most wrong. The danger arrives whether or not you practised for it. The only question is whether their first hour of dark driving happens with you in the passenger seat or alone six weeks after passing.

There is more on the sponsor role, including how to run a session without turning into a second instructor, on our page for parents.

After the test is not the finish line

The report has two findings about the period after licensing that most coverage skipped, and both are about parents. The first is vehicle access: teenagers with primary access to their own car had higher safety-critical event rates than teenagers sharing a family car, and the effect was strongest among the teenagers who had practised least. Gershon and colleagues, in a separate 2018 analysis, found teenagers who owned their car had kinematic risky-driving rates about 60% higher than teenagers sharing one, and they cite earlier work by García-España and colleagues putting the crash risk of teenagers driving their own vehicle at more than double that of teenagers with shared access.

The second is that adult ride-alongs after licensing were rare, and got rarer. In the SHRP 2 data an adult was in the front seat on 17.4% of trips in the first six months of solo driving for the less-practised teenagers and 4.7% for the more-practised, and for the less-practised group that share collapsed to 2.6% in months six to twelve. Alone or with teenage passengers was the norm in both groups. The report's recommendation is simply to keep riding with them after they pass. In Irish terms that maps onto the two-year N-plate period, which is the exact window these findings speak to. So: keep sharing the car for a while rather than handing over a set of keys of their own, and go with them the first time they drive somewhere new. If the own-car question is live in your house, we have written about it directly at buying your teenager a car.

If you are a parent reading this and the honest problem is that you are not confident enough in your own driving to supervise well, that is a common and completely fair thing to say, and it is fixable. Call Kris on 087 394 8102 and ask about a sponsor session, where the instructor takes the learner and the parent out together so everyone leaves practising the same method.