Why does novice drivers’ risk decrease by almost 50% during the first year, and what do they learn? A literature review of explanatory mechanisms

1 Introduction

1.1 Background

Young and inexperienced drivers have on average, a significantly higher accident risk than experienced drivers, both in Norway and in other countries (Elvik et al., 2009; Cassarino and Murphy, 2018; Peck et al., 2008). How much higher the risk is among young drivers than among other drivers varies greatly between studies. Studies from 1982 to 2008 from different countries (Australia, Denmark, France, the Netherlands, the United Kingdom, Sweden, Germany and the United States) summarized by Elvik et al. (2009), show that young drivers have between three and 17 times the accident risk of an average driver. In 2018, drivers in Norway in the age group 18–24 had an average of 3.3 times the risk of drivers aged 45–54 (Bjørnskau, 2020). Factors contributing to the high risk among young drivers include their not yet fully developed cognitive skills and their greater susceptibility to social influence and distractions (Cassarino and Murphy, 2018). For example, young drivers have been shown to overestimate their driving skills and underestimate risk to a greater extent than older drivers.

A review of a number of scientific studies in the Traffic Safety Handbook (2023) shows that the most important measure for reducing the accident risk of new drivers is accumulated driving experience. Two older British studies (Maycock et al., 1991; Forsyth et al., 1995) showed that the risk of new drivers decreases with increasing age by about 10% for each year, but that the risk during the first year of driving decreases by about 40%. Two American studies (Curry et al., 2015, 2017) show that the risk for drivers who start driving when they are 17 years old decreases by about 65% during the first 2 years. The largest reduction in risk occurs in the first 6 months (−35%). In a Norwegian study, Sagberg (1997) found that the risk of accidents among young drivers decreases by almost 50% during the first 10 months after passing the driving test and that the risk continues to decrease during the first 5–7 years of driving.

It is important to understand the reasons why new drivers risk decreases so much during the first year, or what they learn through accumulated driving experience during the first months, so that this learning can be facilitated and reinforced in ways that lead to fewer accidents. Accumulated driving experience refers to the experience gained through regular and continuous driving. In 2023, The Norwegian Public Roads Administration recommends that you practice driving for 140 h before starting your driving test. Some define accumulated driving experience only as the driving experience you get after you have learned the basics, and before you start driving. We define accumulated driving experience broadly here, as all the practical driving experience new drivers get in the time before and after they get their driver’s license. For example, you can get training through private practice driving or at a traffic school before you get your driver’s license, and then you also get accumulated driving experience after you get your driver’s license, when you start driving yourself. What we are concerned with in this study is the practical driving experience of new drivers, which reduces their risk of accidents in traffic. That is why we define accumulated driving experience broadly.

Knowledge of the factors that explain new drivers’ risk reduction can be used to recommend specific measures to strengthen the accumulated driving experience that new drivers obtain from the time they start driver training until they get their driver’s license, for example through increased private practice driving, increased number of driving hours, graduated driver’s licenses, etc. We can imagine that different ways of increasing the accumulated driving experience, for example under safer conditions, can lead to a reduced risk of traffic accidents. Firstly, one could set requirements for increased practice driving before one can get a driver’s license (given that driving with, for example, a parent or a traffic instructor gives a lower risk than when driving alone later). Previous studies indicate that increased practice driving before the driving test will have a beneficial effect on the risk of accidents after the driving test. Gregersen et al. (2003) find, for example, that those who had started practice driving as 16-year-olds had a 46% lower accident risk when they had received a driver’s license without restrictions than those who started when they were 17.5 years old. This is attributed to increased driving experience through practice driving. Secondly, it is conceivable that driving experience can be achieved in a way that reduces the risk of accidents through various forms of graded driving licenses (Belloc and Ivaldi, 1991; DEKRA, 2017; Gulliver et al., 2013; Williams, 2007). Graded driving licenses mean that young people who initially receive a driving license can only drive under certain conditions, e.g., only at certain times (i.e., not in the evening/night or on weekends) and/or not with passengers.

To our knowledge, there are few studies that systematically compare the mechanisms that explain why new drivers’ risk decreases through accumulated driving experience. One of the exceptions is Sagberg and Bjørnskau (2003), who tested three hypotheses that could explain why new drivers’ risk decreases through accumulated driving experience: (1) that technical skills improve, (2) that the perception of hazard improves, and (3) that the understanding of, and ability to interact with other road users improves. Sagberg and Bjørnskau (2003) investigated this in a study of 130 drivers who were recruited to participate in a video experiment and questionnaire survey either 1, 5 or 9 months after the driving test. They find particular support for hypotheses 1 and 2 and partly for the third hypothesis. However, the study by Sagberg and Bjørnskau (2003) is over 20 years old, and there is a need to update knowledge about the mechanisms that cause new drivers’ accident risk to decrease drastically through accumulated driving experience.

1.2 Aims

The aims of the study are to:

• 1. Map the mechanisms that explain why the risk of new drivers decreases during the first year.
• 2. Investigate what the research suggests about which mechanisms are most important in explaining this
• 3. Discuss practical implications, for example how to facilitate the accumulated driving experience to unleash the potential of these mechanisms.

2 Theoretical framework and previous research

2.1 The GDE model

It is relevant to use the so-called GDE model (Goals for Driver Education) (Hatakka et al., 2002) as a starting point for classifying the learning of new drivers. This model forms the basis for driver training in several countries with a high level of road safety, for example Norway, Sweden, Finland. The levels in the GDE model are presented in Table 1.

Table 1. Conceptual focus of the GDE levels and examples relevant to driver education.

Authors’ summary based on Hatakka et al. (2002) and the Norwegian driver-training curriculum (Statens vegvesen, 2016). The fifth level, social environment, was added to the GDE framework after the original four-level model.

The GDE model describes different levels of skills and knowledge that are learned, different factors that affect the risk of accidents and factors that can be included for self-evaluation in driver training. At the first level, you will find “maneuvering the vehicle,” for example driving skills such as braking and turning, appropriate speed, etc. At the next level, you will find “managing traffic situations.” Relevant skills and knowledge here are related to knowing about and following traffic rules, communicating and interacting with others in traffic. The third level is “travel choices and conditions related to travel.” Relevant skills and knowledge here are related to planning the trip, for example destination, route, importance of fatigue, alcohol for driving, etc. The fourth level is “general action and assessment tendencies and ways of seeing the world.” Relevant skills and knowledge here are knowledge and reflexivity related to how one’s own lifestyle, motives, values, health and habits affect one’s own driving style. The last level was added in 2010, later than when the model was first constructed in 1999, and is called “social environment.” Relevant knowledge and skills here are reflexivity related to how norms and peer pressure in the social environment one is a part of influence one’s own driving style.

The basis for the description of the different levels, skills and risk factors at each level is, among other things, a review of accidents involving young drivers (Hatakka et al., 2002). This review highlights, among other things, the importance of risk factors at levels 4 and 5, related to deliberate risk-taking, for example high speed and drunk driving, and links this to subcultures among young people, where risky behavior is “normal” and “accepted.” Hatakka et al. (2002) point out that so-called motivational risk factors (Levels 4 and 5) are a more important cause of accidents than risk factors related to, for example, vehicle handling (Level 1). However, all levels are associated with different risk factors that are important in traffic accidents involving young drivers. It is therefore important to investigate how learning at these specific levels can explain the risk reduction that accompanies accumulated driving experience for new drivers.

2.1.1 The GDE model for accumulated driving experience

The GDE model provides an important overview of the different aspects of driver competence; things one learns as a driver and different levels of learning. The model is important to have as a backdrop, because it is fundamental to the understanding of driver training. However, it is inherent in the model that it takes a long time to learn, or to acquire the knowledge and skills associated with the highest levels (Sagberg and Bjørnskau, 2003). These are reflexive abilities that presumably require maturation, although it is intended that driver training should facilitate this maturation, by initiating reflexive processes in those who participate in driver training.

Sagberg and Bjørnskau (2003) write that it is unlikely that such a large drop in risk that new drivers experience after they have received a driving license can be explained by general social and motivational factors, such as a change in the tendency to take risks, thrill-seeking or other factors that can be particularly linked to lifestyle among some young people. This refers to traffic skills at levels 4 and 5 in the GDE model. Furthermore, Sagberg and Bjørnskau (2003) point out that changes in such conditions must be assumed to occur over a relatively long period of time, e.g. over several years. They therefore assume that when the risk falls by 50% during the first 10 months, it is a learning process that occurs through exposure to different traffic situations. We elaborate on this in the following.

2.2 Three hypotheses about accumulated driving experience

Sagberg and Bjørnskau (2003) test three hypotheses about mechanisms of learning that occur through exposure to different traffic situations. These three hypotheses about mechanisms can be defined at levels 1 and 2 of the GDE models. The three hypotheses that they investigate are:

2.2.1 Getting used to the car

Increasing driving experience makes the handling of the car gradually more automated and error-free, so that the risk of accident-causing errors is reduced. The background to the hypothesis is that an important reason for the high risk of new drivers is that they make mistakes more often because the car they are driving is unfamiliar. This hypothesis fits into a long tradition in road safety research that focuses on various forms of error as decisive for accidents (Parker et al., 1998). A number of studies show that driving an unfamiliar vehicle poses a higher risk, presumably because it is easier to make mistakes with an unfamiliar vehicle, and probably also because this can increase the cognitive load, and thus provide less capacity for other tasks (Lee et al., 2005; Lundberg and Hakamies-Blomqvist, 2003). In the study by Lee et al. (2005), drivers were measured over a longer period of time while they were driving in normal traffic. People who were given a loaner car at the end of the study period had a higher relative risk during the four weeks they drove it, compared to the previous period when they drove their own car. This suggests that it may take some time to gain enough experience with an unfamiliar car for the risk to be reduced.

Sagberg and Bjørnskau (2003) test this hypothesis and find that the assumption that errors related to the operation of the car decrease during the first months is clearly supported, and that the decrease occurs more rapidly for men than for women. This can probably be partly explained by the fact that women drive less than men.

2.2.2 Perception of hazard

Driving experience allows you to learn to identify early signs of possible dangerous situations, giving one more time to take the right precautions. Many studies assume that one of the things that separates inexperienced from experienced drivers is the ability to perceive potentially dangerous situations (Moran et al., 2019). In English, this has been called “hazard perception,” which we can translate as “perception of danger” or risk perception. More specifically, “hazard perception” has been described as the ability to anticipate dangerous situations based on perceptual cues in the situation, or in other words, establishing correct expectations about what may happen (see, among others, Groeger and Chapman, 1996). This is an example of what is called expert knowledge, which is characterized by the ability to quickly and automatically perform the right actions in a number of different situations. Sagberg and Bjørnskau (2003) write that it is assumed that the acquisition of such knowledge consists of establishing an increasingly differentiated set of cognitive schemas, i.e., memory images with which the current situation is compared. As one gains more experience, one becomes better at correctly recognizing more and more situations, and one becomes better at doing this on the basis of increasingly less obvious “signs.” Therefore, dangerous situations will be identified earlier, so that one has more time to act correctly. The schemas are assumed to consist of a set of links between situations and actions; When a situation in a schema is identified, the action follows more or less automatically. As schemas are more or less correct, the actions will be more or less adequate. With more experience, it is assumed that one forms schemas for more situations, so that one acts adequately in more and more situations. The automation means that known dangerous situations are recognized quickly, and that this becomes less mentally demanding. This will free up attention that can be used in other, less familiar situations for which there are no schemas. The inexperienced driver, who has not yet automated the processing of information, must therefore use a disproportionate amount of his mental capacity on driving.

The hypothesis that hazard perception improves during the first 9 months was only partially supported in the study by Sagberg and Bjørnskau (2003). They mention that experience may be particularly important when it comes to quickly perceiving situations that are complex, highly unexpected, or require looking far ahead. They emphasize that further studies based on systematic classification of hazard situations are necessary to shed more light on this. They also conclude that it appears that the capacity to react to traffic situations under additional cognitive load also improves with experience, and that this indicates that inexperienced drivers have less attentional capacity to handle unforeseen situations.

2.2.3 Interaction in traffic

Sagberg and Bjørnskau (2003) write that there may be reason to believe that the high risk of new drivers may be a result of them not having learned the written and especially unwritten rules of traffic. They distinguish between two aspects of interaction competence that it is important to distinguish from each other, namely active and passive interaction competence. Active interaction competence means that you as a road user are able to predict how other road users will act, and that you adapt your own actions accordingly. There may be reason to assume that this is something that is learned with increasing experience. Bjørnskau (1994) shows that drivers who possess such active interaction competence have a significantly lower risk of colliding in traffic. Passive interaction competence means that you as a road user behave in a way that makes it easy for others to predict what you will do. This again means that the more “normal” or “usual” your behavior is, the more predictable you are.

Sagberg and Bjørnskau (2003) find in their study clear changes during the first 9 months for passive interaction, in that the inexperienced drivers quickly acquire a driving style that increasingly resembles that found in very experienced drivers. However, when it comes to active interaction with other road users, relatively little change occurs during this period. The difference between the inexperienced and the experienced drivers suggests that this is a skill that is developed over a longer period of time (Sagberg and Bjørnskau, 2003).

They conclude that the decrease in accident risk in the first months after the driving test is likely explained by a combination of the changes in driving behavior they study.

3 Methods

3.1 PICO form

To specify the research question sufficiently to be able to conduct a search, a PICO schema was developed (see Table 2). PICO is a method for organizing and structuring scientific literature searches. It stands for Population, Intervention, Comparison, Outcome. With the PICO framework, we can clearly define which group we are studying, what kind of intervention or exposure we are looking at, what we are comparing it to, and what outcome we are interested in. This helps us find relevant studies that cover all these aspects and contributes to a more systematic approach to the literature review.

Table 2. Literature search specification (PICO) and examples of search terms.

*No separate search terms for comparison, this was used in the review of studies.

By defining what should be included in the literature review in this way, it also (indirectly) becomes clear what is not included. For example, the population is “New drivers learning to drive a passenger car”. This means, among other things, that studies that focus on refresher courses, training for heavy vehicles or further education of drivers are excluded.

There are several types of outcome measures in the research literature. Firstly, there are outcome measures such as accident involvement, involvement in other traffic incidents, and violations. These are the outcome measures that are best suited to drawing conclusions about the effect on traffic safety, but experience has shown that there are few such studies of training, and even fewer that examine the specific mechanisms underlying the risk reduction over time. This research was summarized in connection with the update of the chapter on driver training in the Norwegian Traffic Safety Handbook (Traffic Safety Handbook, 2023; Høye, 2020; Sagberg and Høye, 2020a; Sagberg and Høye, 2020b; Sagberg and Høye, 2020c).

In this literature review, we have therefore relied on studies with outcomes that are plausible mechanisms, but not accidents or other “hard” measures of road safety. This includes, for example, studies with perception of danger as an outcome measure.

Elaboration of inclusion/exclusion criteria:

• The studies must contain a comparison or relationship that can say something about more and less driving experience/training. For example, compare groups with more and less training, measure accumulated driving experience and include it in an analysis, etc. The accumulated driving experience must not necessarily be obtained before you get a driver’s license; it can be during training or up to a year after - that is, the half year reported by Sagberg and Bjørnskau (2003), plus a little plus a little.
• Evaluation of specific courses, e.g., hazard perception, is only relevant if there is a focus on, e.g., differences in previous competence or prerequisites etc. for those who have trained to varying degrees. In that case, some of these must be fairly new drivers in class B.
• New drivers class B or equivalent must be included, not just young drivers (too imprecise proxy for experience) or courses for advanced drivers.
• The study must investigate the relationship between accumulated driving experience/training (loosely defined) and some type of skills, perceptions, mechanism that is vaguely related to the mechanisms we put forward. It is not sufficient that the study only examines the direct relationship with accidents.

3.2 Search and inclusion criteria

Table 2 was also used as a starting point for search strings. Due to different technical capabilities in the two databases, the search strings were constructed differently for the two locations searched. ScienceDirect has limitations on how many logical operators can be used simultaneously, while ISI has the ability to combine multiple searches. Table 2 shows examples of search terms, and the complete search log is in Appendix 1.

To conduct a thorough and systematic literature review, we established specific inclusion criteria based on the research questions and purpose of our study. These criteria were used as guidelines to assess the relevance and quality of the studies we identified. The inclusion process was based on a comprehensive assessment of the title, abstract, and full text of potential studies.

To be included in our review, studies had to meet the following criteria:

• 1. Focus on the relationship between accumulated driving experience and road safety among new drivers, as defined in Table 2. That is, the study focused on, or contained information about;

• Population: New drivers learning to drive a passenger car, or new drivers
• Intervention: Amount of practice driving (supervisor), hours of practice driving, accumulated driving experience
• Compared: Different degrees of accumulated driving experience, alternative training approaches, new vs. somewhat more experienced drivers
• Outcome measures were mechanisms that can explain the relationship between accumulated driving experience and safety. For example, hazard perception, vehicle knowledge, situational awareness, risk assessment, skills, etc.

• 2. Presents empirical results, i.e., comments and posts without data material are not included
• 3. Be published in English
• 4. Be available in full text and
• 5. Be published in scientific journals or conference proceedings.

We performed a stepwise process to exclude studies that did not meet our inclusion criteria. First, duplicates were removed, then titles and abstracts were reviewed to identify studies that clearly did not fit our criteria. Finally, the full-text versions of the remaining studies were carefully reviewed to determine whether they met the inclusion criteria. Studies that did not meet the criteria were excluded from our final analysis.

The literature search was conducted in June 2023 in two scientific databases: Science Direct (SD) and ISI Web of Science (WOS).

The literature search was conducted in the following steps:

• Database searches. We conducted searches in the ISI web of knowledge (ISI) and Science Direct (SD) databases. Hits on the search terms described in Table 2 were included if they met the following criteria:

• The search terms were found in the title, abstract or keywords
• Published in English
• Published as research articles, review articles, or conference abstracts (not book chapters and reference works)
• Review of titles. Studies from step 1 were excluded if they were clearly not relevant to the “PICO” defined in Table 2. Among other things, studies were excluded if they dealt with truck drivers, professional drivers, older drivers, were only a review of regulations, focused on parental attitudes, etc. This exclusion was made by reading only the titles of the publications.
• Review of abstracts. Studies were excluded if, based on the information in the abstracts, they were not in accordance with PICO, cf. step 2. This exclusion was made by reading the abstracts of the publications.
• Full text review. Studies were excluded if, based on the information in the full text, they were not in accordance with PICO, cf. step 2.
• Compilation of remaining studies. The remaining studies were summarized in tables (see Chapter 4).

We have also included relevant studies that we were aware of from several different sources, for example studies that we were already aware of, or studies that we became aware of by reviewing the reference lists of relevant studies that we had already identified.

3.3 Thematic analysis of results

We have conducted a thematic analysis in the review of the studies in an effort to identify overarching mechanisms that may explain the decrease in risk for new drivers. A thematic analysis is a systematic method for identifying main themes in text material (Braun and Clarke, 2006). In the first step of the process, we read the results from each individual study to classify the studies into overarching codes. The codes were then systematized and arranged into broad categories. In the next step, the resulting categories were reviewed. In this part of the process, we assessed the categories against each other and against the material, and necessary adjustments were made. The purpose of the thematic analysis was to group the studies into more overarching categories of mechanisms.

3.4 Implementation

The searches in the databases Science Direct and ISI Web of Science resulted in 909 and 1,636 hits (excluding duplicates from each database), respectively, which were downloaded. After removing duplicates, 1,693 studies remained that were screened by titles. 1,459 studies were excluded by title screening, while 234 abstracts were reviewed. In the screening of the abstracts, 168 studies were excluded, while 66 proceeded to the next step—review of the full text. Eleven studies could not be found in the full text. The results of the study selection are shown in Figure 1.

4 Results

4.1 Methods, research design, and quality

In this section, we provide an overview of the methods, research design, and quality of the studies we have identified that provide insights into the mechanisms that explain the decline in risk for new drivers during the first year. We include detailed information about each study and identify the mechanisms covered by the studies. We also summarize the main findings reported in the studies to provide a clear overview of the research area. Despite an extensive and detailed literature search, we identified only 18 relevant studies, which were included in this literature review. There are significant differences between the studies both in terms of research methods and associated strengths and weaknesses, and in the topic or mechanism(s) being investigated.

The studies varied in design, including randomized controlled trials, cohort studies, and qualitative studies. Several types of studies followed the same drivers over time: Roman et al. (2015) followed 1,148 drivers for 3 years after they received their driver’s license, using the same questionnaire on different types of driver behavior at 4 different times. In the study by Day et al. (2018), researchers interviewed new drivers 1, 2, and 3 months after they received their driver’s license. In the study by Eshani et al. (2017), a group of students’ first and last 10 h of practice driving were examined using in-car cameras, GPS, audio recordings, and trip recorders. Also, in the study by Durbin et al. (2014), students completed a driving test in traffic at two different times during their training. Durbin et al. (2014) included experienced drivers who completed the same driving test, and examined, among other things, how often the instructor had to intervene, the number of errors in vehicle handling and how many had to interrupt the driving test for safety reasons. The study described by Stavrinos et al. (2021) examines visual attention (use of gaze) in a simulator among young people with different levels of driving experience over time (i.e., at multiple points in time).

Most studies compare different groups at a single point in time. In the study by Boccara et al. (2011), both students and traffic instructors assess the students’ skills, and comparisons are made between students at different stages of training. Møller et al. (2021) compare people who chose a training course with an early driving license for supervised driving (from the age of 17) with people who got a driving license at the age of 18 (no supervision). Stanojević et al. (2022) compare people who got a driving license before and after the introduction of graduated driving training in Serbia.

In many studies, the comparison between groups with different levels of experience is made based on performance on different tasks in a simulator or on a PC/screen. This applies to most studies on hazard perception: In the study by Damm et al. (2011), three groups of drivers are compared; novices who have had early, private driving practice, novices who have not had early driving practice, and experienced drivers. Lazaro et al. (2022) also compares drivers with different levels of experience in a simulator task, and the same applies to Ferguson (2003), Freydier et al. (2016) and Upahita et al. (2018). In the studies by LaVoie et al. (2018) and Scialfa et al. (2011), groups of drivers with different levels of experience are also compared, but in this case, everyone completes a video task.

The quality of the studies varies, and several have significant methodological weaknesses. The literature reviews that summarize large amounts of research are of high quality (Cao et al., 2022; Moran et al., 2019; Robbins and Chapman, 2019; Scialfa et al., 2011), while some of the other studies have very few participants (e.g., Day et al., 2018; LaVoie et al., 2018; Lazaro et al., 2022; Freydier et al., 2016), and/or differences between groups make it difficult to assess which effects are due to different experiences (e.g., Freydier et al., 2016; LaVoie et al., 2018; Damm et al., 2011). The studies that examine student-parent dyads over time (Durbin et al., 2014; Eshani et al., 2017) have no major methodological shortcomings, but since they are students in the training course, they may miss changes that only occur after a longer period of time. Since the included studies are very different, it is difficult to make direct comparisons of methodological quality. Some nevertheless stand out as being particularly methodologically solid, and/or with few major methodological weaknesses. Roman et al. (2015) followed the same young drivers over 3 years and analyze the results in a way that provides a good understanding of how errors and violations develop over time for the individual driver. The study by Stavrinos et al. (2021) examines young people of different ages and degrees of driving experience in order to be able to distinguish the effects of maturation and experience from each other and appears very promising, but since data collection is ongoing, the results are somewhat limited.

There is also great variation in the number of participants the studies are based on. In studies based on questionnaires, the number of participants is typically the highest. For example, the study by Roman et al. (2015) has 1,148 participants, and most of the analyses by Møller et al. (2021) use data from 373 respondents. Several of the studies where students drive in traffic also have relatively high numbers of participants (n = 90 students in the study by Eshani et al., and n = 89 students in the study by Durbin et al.). Several of the simulator studies, on the other hand, have fewer participants, who are often divided into different groups so that the number of, for example, experienced drivers is very low (e.g., the 36 participants in Damm et al., 2011 are divided into three groups of 12 participants).

4.1.1 Study design

As mentioned, there is a wide variation in the research methods of the studies included. Table 3 shows the research designs that were used, and how many participants the included studies are based on.

Table 3. Overview of included studies.

Accumulated driving experience has been investigated using several different approaches. Several studies examine individuals at different stages of driver training, often in combination with a group of more experienced drivers. For example, Boccara et al. (2011) examines groups of drivers at different stages of training, and also a group of more experienced drivers. Eshani et al. (2017) and Durbin et al. (2014) examine the same students early and late in the training course, and Durbin et al. also included a group of more experienced drivers. Some studies compare drivers who have taken training courses that included an early period of limited driving license/supervised driving with drivers without such a period (Møller et al., 2021, Ferguson, 2003, Ferguson et al., 2003). Some studies also examine the number of hours or distance driven (e.g., Eshani et al., 2017; Durbin et al., 2014).¹ In many studies, experience is examined as the time since obtaining a driver’s license in months, and in some cases years (Evans et al., 2022; LaVoie et al., 2018), while Lazaro et al. (2022) combines time since obtaining a driver’s license and how often and how much the drivers drive. In the literature reviews (Cao et al., 2022, Moran et al., 2019, Robbins and Chapman, 2019), the definition of experience varies considerably in individual studies included. A consequence of the different ways of examining experience is that people who are considered inexperienced in one study can in principle be more experienced than both the inexperienced and the most experienced in another study. For example, all participants in the study by Lazaro et al. (2022) have a driver’s license, while the study by Ehsani et al. (2017) is based on people under training. At the same time, we do not have enough information to assess whether the people in the study by Lazaro et al. (2022) have driven more than the individuals in the study by Ehsani et al. (2017) as the latter are in a training course with a lot of mandatory practice driving, while no information is provided about the training of the participants in Lazaro et al. (2022). Similarly, it is difficult, if not impossible, to directly compare the experience across other studies.

4.2 Overview of main findings on mechanisms

We focus on the first objective of the study which is to identify the mechanisms that explain the decline in risk for new drivers during the first year. Table 4 provides an overview of the mechanisms identified and which studies cover each mechanism. Each row represents a unique study, and the columns indicate whether the study covered the specific mechanism or not. The table also contains a summary of the main findings related to each mechanism based on the studies.

Table 4. Mechanisms and main results of the studies.

4.3 Summary of mechanisms

In this section, we focus on the second aim of the study, which is to examine what the research shows about which mechanisms are most important in explaining that the risk of new drivers decreases during the first year. The presentation is based on a thematic analysis of the studies in Table 4, where we group studies that deal with the same more general mechanisms into more general categories.

4.3.1 Better understanding of one’s own abilities and limitations

Studies by Boccara et al. (2011) and Møller et al. (2021) suggest that new drivers develop a better understanding of their own skills and limitations during training, or during their first time as independent drivers. Boccara et al. (2011) compared students’ assessments of their own abilities with corresponding assessments from their driving instructors. As expected, they found that both students’ and instructors’ assessments became increasingly positive over the course of training, but students’ overestimation of their own abilities decreased over the course of training.

Møller et al. (2021) found better skills in terms of handling the car and processing information ²among those who drove for a year with a companion before getting a regular driver’s license, but not better “safety skills,” which are more about the will and ability not to commit offenses and violations. The authors therefore point out that the additional experience that can be gained from a year of accompanied driving may be too small to have clear safety benefits, but that there may also be other types of differences between those who choose different types of driver’s license that contribute to the findings.

4.3.2 Better vehicle handling

The studies by Day et al. (2018), and Ehsani et al. (2017) suggest that interaction with the vehicle becomes more automated and gradually takes less focus. Also, in the study by Møller et al. (2021), people with more experience were better at vehicle handling. These studies indicate that one of the mechanisms that can explain that new drivers’ risk decreases through accumulated driving experience is more automated interaction with the vehicle, so that it does not require much energy and takes focus away from other things. The results from Durbin et al. (2014) also partially support such an interpretation.

In the study by Day et al. (2018), participants initially described how they took the easiest and already familiar routes in order to be able to concentrate on basic vehicle handling, but that this improved during the first few months of driving. After a couple of months, several reported that they drove with more fluidity and handled the car without thinking. Many also stated that this was related to better confidence in their own abilities, and to accepting smaller distances and gaps in different contexts. Participants also experienced increasingly improved situational awareness; at the beginning, several had challenges with knowing where to look or position themselves and uncertainty about how the traffic situation would develop. With more experience, they gained more focus on the overall traffic situation, and became better at generalizing between traffic situations, and anticipating what would happen.

Eshani et al. (2017) also examined driver errors over time. In this study, fewer errors were found over time, indicating better skills. In this study, they did not find a decrease in errors related to vehicle handling (sudden braking, lane management, turning), but they found decreases in both decision-making errors (incorrect speed for conditions, size of the gap, failure to stop/give way) and in errors related to identifying hazards or inattention. The risk of having to interrupt a driving test for safety reasons was lower for people with more driving experience in the study by Durbin et al. (2014). However, the study showed no decrease in safety-relevant critical errors early and late in the driver training course, nor a clear relationship between the amount of practice driving and critical errors in the driving test. Adult, experienced drivers who completed the same driving test, on the other hand, made significantly fewer critical safety errors. Roman et al. (2015) also found increases in driver errors over time, in a measure of driver errors that includes vehicle handling. In the study by Roman et al. (2015), there was no correlation between the amount of driving errors and oversights 6 months after obtaining a driver’s license and how the amount of driving errors and oversights changed in the following months.

4.3.3 Speed control and positioning

Freydier et al. (2016) showed that drivers with less practice driving had greater challenges with speed regulation and maintaining lane position than drivers who had more practice driving, while in the study by LaVoie et al. (2018) drivers with more experience had faster speed judgments than less experienced, younger drivers. Both the study by Freydier and the study by LaVoie are methodologically weak, so the results must be considered uncertain. In the study by Eshani et al. (2017), on the other hand, errors related to lane management increased from the first survey (less experience) to the last survey (more experience).

4.3.4 Risk perception and attention management

Another mechanism that emerges in the studies is that one gains a better developed risk perception, so that one becomes better at knowing what important signals of danger are, and that one becomes better at distributing attention with more accumulated driving experience.

Several studies suggest that the use of gaze and attention becomes better, or more efficient, with more experience. In the study by Evans et al. (2022), experienced drivers were better at detecting hazards related to visibility than drivers who had less experience. Similarly, Stavrinos et al. (2021) show that over time, eye movements change in such a way that drivers gradually reduce irrelevant “scans” of the environment (vertically) and increase horizontal “scans.”

In the study by Scialfa et al. (2011), novices took longer to respond to hazards, even when the authors took reaction time into account. The literature review by Cao et al. (2022) on hazard perception showed that more experienced drivers have shorter reaction times, see hazards faster, and are better at spreading their attention over a wider and more relevant area of the environment. The mechanisms underlying the fact that drivers develop the ability to perceive hazards as they gain more experience, are that they develop knowledge of recognizing signs of hazards, knowledge of precursors to hazards, and become better at spreading their visual attention (Cao et al., 2022). This is precisely what underlies many studies of hazard perception. Upahita et al. (2018) also find similar results; experience is associated with faster reaction time.

The literature review and meta-analysis by Robbins and Chapman (2019) of how drivers’ visual search changes with experience is consistent with the results on hazard perception; novices use their gaze in a more narrow way, while more experienced drivers use their gaze in a broader field.

In the study by Lazaro et al. (2022), experienced drivers had more responsive attention; they reacted to a greater extent with altered levels of attention, and with stress, in the face of dangerous situations. The fact that less experienced drivers did not show the same changes was interpreted as their inability to distinguish equally well between more and less dangerous situations. Damm et al. (2011) found that new drivers who had driven more with a companion positioned themselves better in the lane than other new drivers, but no differences in reaction time or detection of challenging elements.

4.3.5 Other/unplaced

In the study by Roman et al. (2015), a battery of questions is used that covers interaction with other road users, speed adjustment, positioning and risk perception. The various indicators of driver behavior are analyzed together, so that conclusions cannot be drawn about the effects of individual elements, but in this methodologically sound study, the authors find that errors increase over time. This is surprising. Roman et al. (2015) suggest that one reason why their results do not correspond to the general decrease in accident risk for new drivers that occurs during the period covered by the study is that the decrease in accident risk is explained by something other than the types of driver errors covered in the study; an example they highlight is risk perception. Another possibility the authors suggest is that, although driver errors are increasing, it is possible that the errors are being committed in increasingly safer situations as drivers become better at reading traffic and avoiding certain types of behavior in situations where it would be particularly dangerous.

5 Discussion

5.1 What did we find about mechanisms?

The first goal of the study was to map the mechanisms that explain why new drivers’ risk decreases during the first year. We found 18 relevant studies, and we grouped the mechanisms they described into four main categories:

5.1.1 Improved risk perception and Improved attentional control

Eight studies indicate that increased experience leads to better developed risk perception and better attentional control (Evans et al., 2022; Stavrinos et al., 2021; Scialfa et al., 2011; Cao et al., 2022; Upahita et al., 2018; Robbins and Chapman, 2019; Lazaro et al., 2022; Damm et al., 2011). These studies show that experienced drivers detect hazards faster and have shorter reaction times. This is the mechanism with the strongest empirical support in this study.

5.1.2 Better vehicle handling

Day et al. (2018), Durbin et al. (2014), and Ehsani et al. (2017) indicate that new drivers become better at handling the vehicle with increased experience and that this gradually becomes more automated. This frees attention from simpler driving actions.

5.1.3 Better understanding of one’s own abilities and limitations

Both Boccara et al. (2011) and Møller et al. (2021) show that new drivers develop a better understanding of their own skills and limitations during training or at the beginning of independent driving time.

5.1.4 Speed control and positioning

Freydier et al. (2016) and LaVoie et al. (2018) show that experience plays a role in speed control and lane positioning. Freydier et al. indicate greater challenges in less experienced drivers, while LaVoie et al. observe faster speed judgments in experienced drivers. These findings must be interpreted with caution due to methodological limitations.

5.2 What is the most important thing you learn through accumulated driving experience?

The second aim of the study was to investigate what the research says about which mechanisms are most important in explaining that the risk of new drivers decreases during the first year. The results from the literature review indicate that the most important mechanism, i.e., the most important thing one learns through accumulated driving experience, is better developed risk perception and better management of attention. It is this mechanism that has been identified in most studies. This is an example of what is called expert knowledge, which is characterized by the ability to quickly and automatically perform correct actions in a number of different situations. Sagberg and Bjørnskau (2003) write that it is assumed that the acquisition of such knowledge consists of the establishment of an increasingly differentiated set of cognitive schemas, i.e., memory images with which the current situation is compared.

However, we must make the reservation that we cannot assume that improved risk perception is actually the most important mechanism, because: (1) We do not know anything about the relationship between improved risk perception and accident risk for new drivers, (2) The number of studies showing an effect may be due to this being a topic that many researchers enjoy researching, or that it is easier to obtain funding to research, and (3) It is not unreasonable to believe that the mechanisms that explain the reduction in risk are related and that they have the best effect in combination, for example when one gains more experience, one becomes better at handling the vehicle so that one can spend more of one’s attention on hazard perception, while also becoming better at hazard perception. The first two conditions are discussed in more detail in sections 5.6 and 5.7. The last condition is discussed in more detail below.

5.2.1 Improved risk perception appears to be dependent on improved vehicle handling skills

Vehicle handling also requires attention and mental resources, and in line with other forms of learning and practicing skills, the tasks associated with handling the vehicle will also become more automatic as one gain more practice, and thus less mentally demanding. The driver’s mental capacity is a limited resource—there are limits to how many things can be registered, viewed, interpreted and thought about at the same time. Therefore, poorly practiced handling of the vehicle will reduce the opportunities to focus on the overall picture in the traffic. Similarly, more automated vehicle handling will free up more mental capacity that can be used on other important things in the traffic situation. It therefore appears that increased automation of the driving task that results from new drivers gaining more experience leads to a lower degree of cognitive load, and that this provides an increased capacity to focus on other things. Increased experience also seems to mean that one is not “overwhelmed” by all the impressions, and that one becomes better at knowing what is important to spend one’s energy on.

5.3 What exactly is it that you learn from?

The evaluated studies indicate that it is exposure to many different situations that leads new drivers to develop their risk perception and gain a better ability to control their attention (plus that exposure to vehicle handling leads to better vehicle handling). However, the studies do not say anything about exactly what happens when one is exposed to a number of different situations in traffic but based on the theory of hazard perception (Groeger and Chapman, 1996), we can assume that gaining increased experience in traffic is about gaining training in interpreting subtle perceptual cues. We assume that with increased experience one acquires an increasingly differentiated set of cognitive schemas, i.e., memory images, from previous situations with which the current situation is compared. This contributes to one becoming better at controlling attention and detecting hazards, so that one becomes better at quickly and automatically performing correct actions in a number of different situations. Learning of different schemas occurs through exposure to a number of different situations in traffic, which leads to training in interpreting subtle perceptual cues. As experience increases, more and more situations will be recognized correctly, based on increasingly subtle perceptual cues. This means that a potentially dangerous situation is identified at an earlier point in time, which provides more time for correct action. We also assume that the actions to handle dangerous situations are also included in the schemas, which will thus consist of a set of links between situations and actions. This means that when a situation is identified, the action follows more or less automatically. In addition to the rapid recognition of known dangerous situations, the automation of the schemas will result in the release of attentional capacity to handle more unforeseen situations, for which schemas have not yet been established. The new driver is therefore assumed to have to use a disproportionate amount of his mental capacity on driving, because information processing is less automated than is the case for the experienced driver.

5.4 What did we not find about mechanisms?

5.4.1 Interaction

A mechanism that is not highlighted in any of the studies we review is that one gains a better developed ability to interact with others. This means that one (a) becomes better at driving in a way that makes others understand what you are doing and does not misunderstand you in ways that can lead to dangerous situations and (b) one becomes better at understanding what others are doing in traffic, so that misunderstandings that can lead to dangerous situations are less likely to arise. Sagberg and Bjørnskau (2003) refer to the first as passive interaction and the second as active interaction. Their study indicates that passive interaction is an important mechanism.

None of the studies reviewed explicitly mention interaction in traffic. However, it can be argued that the studies that show better positioning and speed adaptation in new drivers with increased training can be categorized as improvements in passive interaction. The studies that deal with risk perception and perception of danger can be described as studies that deal with active interaction, since the dangers in traffic are largely other road users. Similarly, several frequently used questionnaires on driver behavior include questions about interaction with other road users, but the results are usually analyzed in such a way that interaction cannot be isolated. This applies, among others, to the “driver behavior questionnaire” used by Roman et al. (2015).

5.4.2 Factors at levels 3, 4, and 5 in the GDE model

The four mechanisms that we identified in the literature study primarily deal with knowledge and skills at levels 1 and 2 of the GDE models: better technical handling of the vehicle (Level 1), better speed regulation and positioning (Level 1) and better developed risk perception and attention management (Level 2). The identified studies say little about factors at levels 3, 4 and 5 of the GDE models. It is uncertain whether this is because no studies have been conducted that say anything about development at these levels, or whether there is little or no development at these levels through accumulated driving experience. However, some of the identified studies also show the importance of learning at higher levels in the GDE model. The studies by Boccara et al. (2011) and Møller et al. (2021) suggest that new drivers develop a better understanding of their own skills and limitations during training, or during the first time as independent drivers. Boccara et al. (2011) found that students’ overestimation of their own abilities was reduced over the course of their education.

Sagberg and Bjørnskau (2003) write that it is unlikely that such a large drop in risk that new drivers experience after they have received a driving license can be explained by general social and motivational factors, such as a change in the tendency to take risks, thrill-seeking or other conditions that can be particularly linked to lifestyle among some young people. This refers to traffic skills at levels 4 and 5 in the GDE model. Furthermore, Sagberg and Bjørnskau (2003) point out that changes in such conditions must be assumed to occur over a relatively long period of time, such as over several years. They therefore assume that when the risk falls by 50% during the first 10 months, it is a learning process that occurs through exposure to different traffic situations. Our study seems to support this.

5.5 Practical implications

The third aim of the study is to discuss practical implications, based on the results of the literature review, for example how to facilitate accumulated driving experience in order to exploit the potential of the studied mechanisms. Our study indicates that the most important thing one learns through accumulated driving experience is to be exposed to a lot traffic situations, so that one gets updated schemas for risk perception. Then one becomes better at anticipating dangers and interpreting situations. In addition, extensive driving practice is important, as it improves one’s ability to handle the vehicle. This is important for automating the driving task and developing “expert knowledge.” Given that the risk decreases a lot in a short time (for example 50% in 10 months), it is relevant to discuss whether accumulated driving experience can be achieved through safe frameworks, i.e., contexts that involve lower risk than when driving as an independent driver. In this discussion we base ourselves in particular on results from previous studies, which are reported in the Traffic Safety Handbook (2023).

5.5.1 Is increased requirements for practice driving a good idea?

The first example of a safe framework is practice driving. Previous studies indicate that an increased accumulated driving experience before the driving test will have a beneficial effect on the risk of accidents after the driving test. Gregersen et al. (2003) find, as mentioned, that those who had started practice driving as 16-year-olds had a 46% lower risk of accidents when they had received a driver’s license without restrictions than those who started when they were 17.5 years old. This is attributed to increased training through practice driving. Informal training can reduce the risk of accidents. Longer periods of informal (peer-supported) training have been shown to be more effective than shorter periods (Sagberg and Høye, 2020b). This suggests that increased requirements for practice driving are a good idea from a traffic safety perspective. In 2023, The Norwegian Public Roads Administration recommended 140 h of practice driving before getting a driver’s license. We will not say anything about a specific number of hours, but it seems sensible to implement measures that motivate increased driving practice.

5.5.2 Is a graded driver’s license a good idea?

It is also conceivable that the accumulated driving experience can be achieved in a way that reduces the risk of accidents through various forms of graded driving licenses (Belloc and Ivaldi, 1991; DEKRA, 2017; Gulliver et al., 2013; Williams, 2007). Graded driving licenses mean that young people who initially receive a driving license can only drive under certain conditions, e.g., only at certain times (i.e., not in the evening/night or on weekends) and/or not with passengers. Graded driving licenses are a driving license system that exists in the USA, Canada and Australia, among others. To obtain a driver’s license, new drivers must go through a training phase that mainly consists of companion-supported driving in private, and an intermediate phase where the driver can drive on their own, but with a number of restrictions (such as a ban on driving at night or with passengers). Such schemes have resulted in relatively large accident reductions in the countries where they were introduced, usually as a replacement for far less restrictive schemes. The Traffic Safety Handbook (2023) summarizes the studies by stating that the accident reductions are probably mainly due to reduced exposure, but the risk may also have decreased because young drivers have the opportunity to gain driving experience under relatively “safe” conditions. These results, combined with the results from our literature review, suggest that graded driving licenses can also contribute to increased training in “safe” environments.

5.5.3 Accumulated driving experience using digital tools?

In their literature review, Roman et al. (2015) refer to studies that suggest that young drivers’ attentional use, more specifically perception and reaction to hazards, situational awareness and avoidance of distractions, can be improved by using computer-based training programs, including over the internet. Cao et al. (2022) have conducted a more recent, more comprehensive review of research on hazard perception (which is not specific to young drivers). According to Cao et al. (2022), it has been documented that hazard perception training programs have an effect on the typical behaviors measured in such studies; shorter reaction times, more potential hazards being detected, and better use of gaze (more dispersed and anticipatory “scanning”). Cao et al. (2022) finds that a combination of training approaches, such as instruction, expert demonstrations, and active practice with feedback, can improve these behaviors. Other studies have shown that drivers who fail or perform poorly in such tests have a higher risk of accidents than drivers who pass the test (see Sagberg and Høye, 2020a). However, it is not clear whether existing training measures provide lasting changes in knowledge in such a way that it will affect driving in traffic (Cao et al., 2022). In addition, it is important to mention that we assume that the combination of experiences gained through driving yourself is important, i.e., both exposure to the vehicle and the traffic. We can hypothetically imagine that one can become better at-risk perception in traffic, for example through PC games and videos. In order to be able to use this knowledge in real traffic, vehicle handling must be automated, so that one has the capacity to perceive danger. This means that digital tools must simulate the entire driving process as realistically as possible.

Blom et al. (2023a) and Blom et al. (2023b) discuss the extent to which new drivers’ accumulated driving experience can be carried out in a simulator, in a way that provides the same learning effect as independent driving on the road. A relevant question then is whether one learns about one’s own limitations under difficult conditions in a simulator, in the same way as one does in accumulated driving experience in a car under real conditions. Blom et al. (2023a) and Blom et al. (2023b) point out that this may depend on the severity of the consequences. A fundamental mechanism here is the students’ perceived risk (Pless, 2016; OECD, 1990), and it will probably be higher for students who drive under real conditions than for students who drive in a simulator. One probably experiences a higher degree of risk if you lose your grip on a real road than if it happens in a simulator. Similarly, experiencing being surprised by things you have not seen will probably affect the risk experience (and learning?) more in a car on the road than in a simulator. Our assumption in the literature review is that through experience with traffic, you develop a large number of cognitive schemas, which consist of links between situations and actions. As experience increases, you will be able to correctly recognize more and more situations, and on the basis of increasingly subtle perceptual cues. It is obvious to conclude that this seems to be the most important thing you learn in accumulated driving experience. Any simulator training must be adapted to this. This means that you must not be trained in particularly dangerous situations in a simulator (e.g. facing wild animals, or swerving trucks in the road, as indicated in some examples). This can lead to excessive belief in your own skills in a way that leads to a dangerous driving style (for example, higher speed). The simulator must instead expose road users to an enormous number of different and more or less dangerous situations, so that a high number of cognitive schemas are in the “experience bank.” These are important questions for future research.

The literature suggests that digital interventions are most promising when they combine active practice, exposure to varied traffic situations, expert feedback, and hazard perception training in realistic environments (Cao et al., 2022; Petzoldt et al., 2013; Roman et al., 2015). Such interventions should primarily support the development of cognitive schemas, anticipatory hazard detection, and efficient attentional management, rather than focusing on performance or risk-taking under extreme conditions. It is also important that digital training complements supervised on-road driving experience, rather than replacing it entirely. A key challenge for future research is identifying how digital tools can provide effective learning experiences without contributing to unrealistic confidence in one’s own driving abilities.

5.6 Methodological weaknesses

Despite an extensive literature search, we identified relatively few studies of good quality. Among the studies included, the quality varies, and several have significant methodological weaknesses.

Are observed differences and relationships in the studies reliable? Several studies have methodological weaknesses that make it difficult to say what is the reason for observed differences or relationships. A common challenge is that people with more and less experience also have different ages, so that the effects of age and experience cannot be separated in a good way. This applies, for example, to LaVoie et al., 2018, Damm et al. (2011), Durbin et al. (2014), Evans et al. (2022), and Lazaro et al. (2022). A related challenge is, as in the study by Møller et al. (2021), that there are other types of differences between people with different degrees of experience, such as interest in driving.

To what extent is the outcome measure relevant to the mechanism? And can the studies be compared? The narrowing of focus that is often required to conduct meaningful scientific research often also means that the conclusions that can be drawn are somewhat limited. An example is when driver errors are measured with questionnaires; then the results will naturally only include those driver errors that are asked about (and that the respondent also remembers and are willing to admit). Whether one can then draw conclusions about driver errors in general depends, among other things, on whether the errors that are investigated are the most common or most important that occur in driving in traffic. This type of difference may be one of several possible reasons why some studies that focus on the same phenomenon report different results. Furthermore, the degree of training, or experience varies somewhat between studies, so that drivers who can be described as experienced in one study (e.g., at the end of the training course with a lot of practice driving) correspond to the most inexperienced in another study.

Can the results be transferred to road driving? For risk perception, many studies show a relationship between driving experience and the ability to perceive danger (Cao et al., 2022), and other research has shown that people who do poorly on risk perception tests have a greater risk of being involved in accidents. This clearly indicates that risk perception is important for road driving, but even though risk perception can be influenced by relatively short training programs, it is somewhat uncertain whether the effects of such training programs are long-lasting, and whether they can influence driving in traffic (Cao et al., 2022).

5.7 Questions for future research

5.7.1 To what extent are the mechanisms actually related to a reduction in risk?

The studies reviewed in this literature say something about what new drivers become better at in the first months, but do not cover whether these mechanisms lead to fewer accidents for new drivers who undergo accumulated driving experience, as the studies do not necessarily include accidents and risk as outcome measures. Previous literature reviews show relationships between risk perception and accident risk (Sagberg and Høye, 2020a; Sagberg and Høye, 2020b; Sagberg and Høye, 2020c). There are also studies that suggest that experience with vehicles is related to accident risk (Lee et al., 2005; Lundberg and Hakamies-Blomqvist, 2003). However, the extent to which these mechanisms explain the reduction in risk in the first few months after obtaining a driver’s license is somewhat uncertain. The only thing we can say something about in our study are the factors that research shows that people improve through accumulated driving experience, and we speculate that these may help explain the decrease in risk for new drivers in the first months. However, we need specific studies to investigate which factors explain the decrease in accident risk. This is an important area for future research. The study that finds that some errors increase over time (Roman et al., 2015) is a good example of the weakness of our design, which only looks at the coincidence of trends over time. However, it is possible that these are errors that are not critical for road safety; we know that the risk for new drivers decreases drastically.

5.7.2 The combination of mechanisms

Our study suggests that the combination of mechanisms is important. For example, we assume that improved vehicle handling is a prerequisite for improved risk perception, because it does not help to be good at perceiving danger if vehicle handling requires most of one’s capacity. We can therefore imagine that there are many mechanisms that work together to explain the decrease in risk for new drivers. Mapping this combination of mechanisms and how they work together is an important area for future research.

5.7.3 Which mechanisms are important when?

Previous research shows that the risk of new drivers decreases over time; mostly in the first months, but also steadily for five to 7 years. We have focused particularly on the first year in this study and identified mechanisms that explain this. These are particularly mechanisms at levels 1 and 2 in the GDE model. However, it is conceivable that other mechanisms are important for new drivers over time. Some of the mechanisms concern age and brain development over time, for example cognitive skills, and that one as a young driver is more susceptible to social influence and distractions, overestimating one’s own driving skills and underestimating risk. (Cassarino and Murphy, 2018). Other factors may result from maturation with increasing age, for example conditions related to levels 4 and 5 in the GDE model, related to reflection on how the social environment affects one’s own driving style and choices in traffic. It is not unreasonable to assume that these are reflexive abilities that can be developed over time, and perhaps also relatively early, through driver training. It is also not unreasonable to assume that young people today mature earlier than young people 20 years ago, for example due to cultural changes related to a greater focus on self-reflection. Perhaps such factors could also help explain the decline in risk for young drivers over time (Bjørnskau, 2020). However, this is speculation, which needs to be investigated in future research.

5.7.4 Are our results due to publication bias?

Our review provides greater support for the idea that development in risk perception is an important mechanism, compared to the study by Sagberg and Bjørnskau (2003). It is relevant to ask to what extent our main result on the importance of risk perception is due to publication bias. Publication bias is often referred to as a tendency among researchers, research institutions and editors to only present publications that show the expected (or desired) effect. This often means that research that shows positive results is published, while publications that show no correlation (negative results) are not published. In our case, it can be hypothesized that risk perception is a popular topic among researchers and those who fund research, and that this explains why we find so many studies on it (and conclude that improved risk perception is an important mechanism that explains what is learned through accumulated driving experience).

5.7.5 How are the demands on drivers’ skills and the effects of accumulated driving experience affected by vehicles that are increasingly automated?

The results of the literature study related to the most important things new drivers learn through accumulated driving experience are mainly based on the vehicle technology we have had over the past 10–15 years. However, the driver’s role will change in the coming years, with increasing levels of automation. This may also change the demands on drivers’ skills, and what is learned through accumulated driving experience. If larger parts of the driver task are automated, it is conceivable, for example, that some aspects of vehicle handling will become less important. For example, using a manual transmission already seems to be less important in a Norwegian context. We can expect that several of the tasks at level 1 of the GDE model will be taken over completely or partially by automated technology in the coming years. This may also apply to some tasks related to risk perception (level 2 of the GDE model), when various sensors and emergency braking technology are further developed. In this context, we can speculate whether skills at level 3 of the GDE model (avoiding distraction and dozing off or falling asleep) will become more important. Perhaps the requirements for drivers will become more similar to those for pilots in aviation, who operate largely with automated systems. This follows from the fact that the driver’s role may change more to a control room operator role, with increasing levels of automation. The importance of avoiding distraction has already been highlighted due to new vehicles, which have large screens that can be distracting. It is also conceivable that some of the technological changes lead to drivers adapting to their behavior and taking higher risks, so that the measures do not lead to lower risks. We have seen a number of examples of this in the past. It is also possible to discuss what happens to skills at levels 4 and 5, as the vehicle fleet becomes increasingly automated. With increasing levels of self-driving technology, it is conceivable that GEO-fence speed limits could be built in for different geographical areas and override drivers’ motivation to drive above the speed limit. Alternatively, it is conceivable that motivational factors (levels 4 and 5 in the GDE model) continue to be important, because drivers can choose to disable such technology. These questions show that future vehicle technology offers a number of important questions for future research, related to driver requirements, risk development and how driver training can be adapted and changed in line with technological developments.

5.7.6 Digital tools, AI, and simulator-based learning

Digital tools, simulator-based learning, and AI-supported driver assistance systems represent important areas for future research related to safe accumulation of driving experience. Recent studies have examined the potential of simulation-based hazard perception training (Javanbakht and Mirbaha, 2024) and AI-supported safety systems such as fatigue detection technologies (Dewi et al., 2026). Such approaches may provide opportunities for exposing novice drivers to a broad range of traffic situations in relatively safe environments. At the same time, an important challenge is ensuring that digital interventions support realistic risk perception, attentional management, and development of cognitive schemas without contributing to overconfidence or risky driving behavior. Future research should therefore investigate how digital training tools can best complement supervised and real-world driving experience.

5.7.7 What is experience?

We define accumulated driving experience broadly. However, not all forms of driving experience are necessarily equally informative or beneficial for safety development. The learning value of experience may depend on factors such as supervision, feedback, variation in traffic situations, task complexity, and whether the experience contributes to the development of hazard perception, attentional management, and realistic self-assessment. In the reviewed studies, accumulated driving experience includes several qualitatively different forms of learning, including supervised practice driving, independent post-licensure driving, simulator-based exposure, and different traffic environments and task demands. This suggests that “experience” should not merely be understood as exposure or hours driven, but also as the quality and diversity of learning situations to which drivers are exposed. Exposure alone may therefore not necessarily lead to safer driving if it does not support the development of cognitive schemas, anticipatory hazard perception, and adaptive attentional strategies. Future research should investigate which forms of experience contribute most effectively to long-term safety development, and how different combinations of supervised practice, independent driving, and digital training environments influence learning outcomes and crash risk reduction.

Future research should also investigate how familiarity with the road environment contributes to the development of hazard perception, attentional management, and anticipatory driving behavior among novice drivers. Repeated exposure to specific traffic environments may help drivers develop cognitive schemas related to route characteristics, typical hazards, and relevant environmental cues. Route familiarity, landmark recognition, and wayfinding processes may therefore influence how drivers allocate attention and anticipate risks in traffic situations. However, these mechanisms remain relatively underexplored in research on accumulated driving experience and novice driver learning.

6 Conclusion

The results of the literature review indicate that the most important mechanism, i.e., the most important thing learned through accumulated driving experience, is improved risk perception and improved management of attention. This is the mechanism that has been identified in most studies. The other mechanisms we find in the literature review are improved vehicle handling, improved understanding of one’s own abilities and limitations, and improved speed regulation and positioning. We discussed various measures that can provide accumulated driving experience under safe conditions, for example through private practice driving, graded driving licenses and digital tools.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. During the preparation of this work, the authors used ChatGPT to check and improve language, and to identify repetitive or overly detailed text that could be shortened or removed. ChatGPT was also used to translate relevant text from Norwegian into English. All AI-generated output was reviewed and revised by the authors, who take full responsibility for the content of the publication.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frsc.2026.1859177/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Mingxiang Feng, Wuhan Planning and Design Institute, China

Reviewed by: Gang Xue, Guangdong Polytechnic Normal University, China

Navid Javanbakht, Imam Khomeini International University, Iran

Footnotes

However, such a detailed measure of experience is not used in the analyses that are relevant for this summary.

Perceptual-motor skills in the “driver skill inventory” (Lajunen and Summala, 1995).