Modelling the Effect of Mobile Phone Use on Driving Behaviour Considering Different Use Modes
Mobile phone use while driving is a major cause of driver distraction, affecting driving performance and increasing accident risk. Governments have responded to this with the implementation of legislation prohibiting the use of mobile phones, under specific conditions, mainly adopting the hands-free use. Still, mobile phone is a cause of several types of distraction rather than just manual. This study explores the effect of mobile phone use while driving via a simulator experiment. Participants drive under various types of mobile phone use mode- namely, handheld, hands-free (wired earphone), and speaker to capture this effect. Results highlight the effect of mobile phone use, regardless of the use mode, on driving behaviour through specific indicators: maximum driving speed, reaction time, and lateral position. In particular, considering the aforementioned parameters the handheld mode demonstrates safer driving behaviour compared to the speaker mode. The results of this study stress the need for a reconsideration of the present legislation.
Mobile phone use while driving is a major cause of driver distraction, affecting driving performance and increasing accident risk. Governments have responded to this with the implementation of legislation prohibiting the use of mobile phones, under specific conditions. In particular, in Europe in most countries mobile phone use is permitted under the hands-free mode; that is, no handling of the device is allowed and the driver should not touch the mobile phone at any time. Hence, the allowed mobile phone use modes may include speaker mode, use of wired earphones, or Bluetooth. France and Spain permit mobile phone use only under the Bluetooth mode. At the same time, Sweden, a country which has adopted the Vision Zero accidents, only recently (February 2018) prohibited the use of mobile phones while driving. Greece follows the hands-free legislation, allowing mobile phone use only with Bluetooth or on speaker mode. According to traffic police data 24.127 violations of mobile phone use while driving were recorded in 2017 in Greece, comprising a proportion of 5,5% considering all recorded traffic code violations of that year. Still, this number is not representative of the actual number of drivers using their mobile phone in an illegal mode when driving; it is rather an indication of infrequent police monitoring. The VINCI Autoroutes Foundation media release  based on a large scale study (sample of 1004 Greek drivers) stated that 55% of Greek drivers violate the traffic code considering mobile phone use, noting that this is the highest proportion considering the 11 European Union countries where the survey was conducted. At the same time, distraction caused by mobile phone use while driving was responsible for 24% of the road accidents with casualties in 2017 (considering the accidents to which an identifiable driver error or violation was attributed), according to the Greek traffic police records.
Still, the adoption of the hands-free legislation in most countries indicates that relevant bodies consider the resulting physical distraction attributed to handling the mobile phone rather risky, while at the same time they do not seem to consider the cognitive distraction as important. Yet, driving is a complex perceptual and cognitive task and as such the effect of mobile phone use is significant even when physical contact is not involved . Strayer et al.  emphasized this effect comparing it with DUI at a BAC of 0.8.
Interestingly enough, mobile phone use while driving affects driving performance considering all three levels of the driving task: operational, tactical, and strategic , and numerous studies have investigated these effects on driving behaviour. The most commonly explored indicators of driving behaviour are driving speed and reaction time. Considering driving speed, the use of mobile phone while driving results in a reduction of the maximum, mean and median speed and an increase of speed deviation [2, 5–15]. Haigney et al.  observed compensatory behaviour with drivers driving at higher speeds after call compared to those before call, a phenomenon which may increase the risk attributed to mobile phone even more. Reaction time increases with mobile phone use [5, 6, 12, 13, 15–20]. In addition to the increased mean reaction time an increase in the minimum, maximum, and standard deviation values has also been observed. Acceleration and deceleration have also been investigated but at a lesser degree, with no clear emerging patterns. Saifuzzaman et al.  observed an increase in mean acceleration while Papadakaki et al.  a reduction with mobile phone use. At the same time Saiprasert et al.  noted a reduction of acceleration deviation. In addition, the effect on normal car-following deceleration and sudden deceleration at event occurrence differed considering various use modes and road network types [6, 9]. Headway, both time and space, comprises yet another explored indicator [2, 14, 17, 21]. Modification of driving behaviour through driving speed, reaction time, deceleration, and headway affects time-to-collision (TTC) which is an indicator of accident occurrence. Lamble et al.  explored time-to-collision for three scenarios: control, phone dialing, and a cognitive task. The latter involved the highest TTC values.
In addition to the longitudinal dimension of driving behaviour, the lateral is also affected with mobile phone use. The most commonly explored parameter is lateral deviation, and results indicate that mobile phone use results in its increase [2, 5, 19, 24]. Garrison and Williams  on the other hand observed a reduction of lane deviation.
The effect of mobile phone use on driving behaviour is also dependent on several exogenous and endogenous parameters. Exogenous parameters include type of road, that is, whether it is urban, rural, or motorway [6, 12], speed limit , prevailing traffic flow , weather conditions , time of day , driver age [17, 27], and gender . Endogenous parameters include conversation complexity [5, 7, 13, 18, 28] and use mode, that is, whether it is handheld, hands-free (use of wired earphones), Bluetooth, or the speaker mode. Mobile phone use familiarity while driving was also found to influence the resulting impact. In particular, Tokunaga et al.  observed improved performance considering the subjective mental workload when talking on the mobile phone while driving for experienced users (i.e., drivers who frequently use their mobile phone while driving).
The present paper investigates the effect of mobile phone conversation on driving behaviour of Greek drivers considering several internal and external contributory parameters. The tool used for the study is a driving simulator. A considerable driver sample size is employed in the experiment, allowing for the correlations between the different variables. In addition to the already explored variables mobile phone specific variables are also considered including familiarisation with the mobile phone and mobile phone use frequency while driving. Further, while exploring the use of several mobile phone use modes, that is, handheld, wired earphones, and speaker, this research also attempts to answer the question of whether the existing legislation on mobile phone use while driving is valid. The paper is structured as follows: in the following section the experimental procedure is described, while in Section 3 the methodology is presented. Results are presented in Section 4, and the main findings are discussed in Section 5.
2. Experimental Procedure
The tool employed for this study was the FOERST Driving Simulator FPF, consisting of a driving seat and 3 LCD 40′′ screens (Figure 1), located in the premises of National Technical University of Athens.
The experiment involved a control (nondistracted) drive and three distracted drives where participants were engaged in mobile phone conversation under three different use modes: handheld (HH), hands-free (wired earphones) (HF), and speaker (SP). The control drive was embedded in a single drive together with a distracted drive. Two road networks were simulated: an urban and a rural. In total, each participant drove 6 drives of about 5 kilometers each. Conversations involved everyday topics and were of medium complexity, and participants answered the phone at the start of each distracted part of the drive. To capture reaction time a stop sign was simulated to appear at the windscreen (on the centre LCD screen), at which drivers were instructed to react by bringing the car to an immediate and complete stop. The order of the distraction scenarios, road network types, location in the network, and position on the windscreen of the stop sign appearance differed both within and between participants in order to avoid order effects. A familiarization drive was also performed for participants to become accustomed to the simulator environment and the vehicle controls. Participants were asked to fill in a questionnaire to collect information on mobile phone use, driving behaviour, and socioeconomic characteristics.
The experiment took place in October and November 2016, and the only quota applied to potential participants (recruitment process involved social media and snowball sampling techniques) was acquiring a valid driving license and owning a mobile phone. Following the experiment, one participant was excluded from the sample as the exhibited driving performance was rather poor; questionnaire data also showed that this participant did not drive at all in her everyday life. The final sample consisted of 50 healthy adult volunteer drivers, 32 males and 18 females, aged between 20 and 60 years old (average of 31 years). For 41 participants it was the first time to participate in a driving simulator experiment, 5 had participated once, and 4 more than once. Drivers’ characteristics are illustrated in Table 1.
Interestingly enough, a substantial proportion of the participants (76%) use their mobile phone while driving, while only 20% acquire a Bluetooth, with the remainder 56% using it in an illegal manner. At the same time, only 1 driver had a complete knowledge of the legislation describing mobile phone use when driving, 25 drivers (50%) had partial knowledge, and 24 (48%) stated they are not aware of what the relevant legislation states. Figure 2 demonstrates the frequency of use in the different road environments and Figure 3 the level of safety that drivers perceive when using it while driving.
Results indicate that drivers use their mobile phone less in rural networks. In addition, Bluetooth mode is the least used regardless of the road network type, while handheld mode and speaker mode are used most in urban and rural areas, respectively.
Only a small proportion of drivers (10% and 16% for urban and rural road networks, resp.) do not feel safe when using their mobile phone while driving. Drivers were asked to define the ways in which their driving behaviour is modified under mobile phone use, to capture the perceived effects. 45% noted that they drive more carefully, 42% reduce their speed, 3% stop the car, 3% drive on the right lane, and 8% stated that their behaviour remains unchanged.
In this study the effect of mobile phone use on driving behaviour is considered through the following parameters: maximum driving speed, reaction time, and lane deviation (standard deviation of lane position). Initially, ANOVA tests were performed for different driving behaviour indicators to test whether mobile phone use affects them. Reaction time (F=2.292), maximum speed (F=2.510), and lane deviation (F=6.626) were found to be affected by mobile phone use, while mean driving speed (F=0.813) was not.
In most studies, driving indicators are treated as continuous parameters where even a small change, increase or reduction, is modelled as a change. Still, specific indicators of driving behaviour such as speed, acceleration, deceleration, and headway are “chosen” by the driver and fluctuate around the desired values (in car-following behavior, speed, acceleration, deceleration, and safety distance are employed in the form of desired values in widely used models ). This indicates that small changes should not be treated as actual changes in driving behaviour. In addition, small changes in these parameters are not expected to significantly change the risk level of driving. At the same time, it is quite common to classify driving behaviour indicators such as speed, reaction time, headways, acceleration, and deceleration, using specific threshold values considering road safety [32, 33]. Hence in this research the parameters describing driving behaviour, rather than being treated as continuous parameters, were classified under three categories and were treated as discrete variables. One disadvantage of this method is that small changes especially in values close to the defined thresholds are allocated to different categories. However, this is a common issue in clustering methodologies.
The three categories were initially formed based on the distribution of the variables between the drivers. The lower and upper 25% (1st and 4th quartiles) formed the low and high category, respectively, while the remainder 50% (2nd and 3rd quartiles) the medium category. In addition to this, the shape of the distribution was also considered (local maximum values, changes in slope) as well as typical values of the explored variables that may denote low, medium, and high values. In particular, for driving speed, speed limits and the type of simulated road were also considered when defining the classification categories. Considering reaction time, several values have been noted in relevant studies but vary considerably between expected and unexpected situations. Threshold values do not exist, but typical values may range from 0.7 to 1.8 seconds [34–36]. Lastly, considering lane deviation neither threshold nor typical values are noted. Hence, classification of this parameter was based solely on the recorded values distribution. The classification of the explored parameters into the distinct driving behaviour categories is presented in Table 2.
Following the classification, discrete choice analysis was performed to the collected data. In particular, ordered probit models with random effects were estimated to exhibit the influence of mobile phone, as well as, other variables on driving behaviour. The dependent variable was represented by the three aforementioned categories low, medium, and high; hence the ordered specification was selected. The random effects portion was added to capture the correlation between the drives of the same individual . The probability of a driver choosing to employ speed at a Y level (low, high, or medium) is dependent on x parameters/explanatory factors and the relevant relationship iswhere Φ is the cumulative normal function, θ_0=-∞<<⋯<θz=∞ are the breakpoints, x is the vector of the explanatory factors, and β is the vector of the unknown parameters.
With the F(y) distribution following the distribution of random errors ε is pictured in Figure 4, with k1 and k2 denoting the thresholds between the three variable categories.
This section presents the estimated models for the explored variables of maximum speed, reaction time, and lane deviation.
Maximum Speed. Driving speed is a crucial factor considering road safety as higher speeds result in not only higher accident risk but also higher severity rates [37, 38]. Findings from existing studies indicate a reduction of the maximum speed when using a mobile phone. Table 3 presents the probit model results for maximum driving speed.
Maximum driving speed is reduced when talking on the mobile phone, with the handheld mode exhibiting the highest reduction, followed by the speaker mode. Increased driver familiarity with the mobile phone results in higher maximum speeds. In addition, drivers who use their mobile phones while driving in rural road networks also exhibit higher maximum speeds. Other contributory parameters considering maximum driving speed include type of road network, attitudes towards driving, age, and gender. In particular, drivers exhibited lower values of maximum speed in the urban environment compared with the rural, participants who love driving (moderately, very, and extremely) exhibited higher driving speeds than drivers who do not love driving, drivers older than 54 years of age exhibited higher speeds compared to drivers aged between 18 and 24, and women exhibited lower driving speeds than men.
Reaction Time. Considering reaction time it should be noted that a “standard” or “generally accepted” reaction time (referred to also as perception-reaction time) cannot and does not exist . Green  suggests five principle variables for reaction time, namely, expectation, urgency, age, gender, and cognitive load. In the designed experiment expectation is similar for all drivers as they are told that a “stop” sign will appear on their windscreen; of course the manner in which they handle this information is subjective. Urgency is also the same, as they are instructed to immediately stop the vehicle. The effect of the three remaining categories shall be explored, with the cognitive load being mainly attributed to the conversation on the mobile phone.
Reaction time is yet another driving behaviour indicator, closely linked to accident risk that is affected by mobile phone use. The highest effect is observed with the use of the speaker mode, followed by the hands-free mode (Table 4). Results indicate that the less the impedance on the driver is, the less alert the driver is; this is in line with findings of Consiglio et al. . Participants who use their mobile phone while driving demonstrated lower reaction times. Increased reaction times were also observed in urban areas probably as a result of the more complex driving conditions demanding for higher workload and the more complex driving environment which increases driver distraction. Other contributing parameters included maximum driving speed (drivers employing higher driving speeds exhibit lower reaction times), accident history (drivers involved in a traffic accident during the past three years exhibit higher reaction times), age (drivers aged 25-34 exhibit lower reaction times than younger drivers aged 18-24), and gender (women exhibit higher reaction times than men).
Lane Deviation. Lane deviation is also an indicator of driver distraction, considering the lateral dimension of driving. The data used for the design of the probit model involved only the rural area, as several road segments in the urban area had two traffic lanes allowing vehicles to make lane changing manoeuvres. Hence, the standard deviation of vehicle position in the lane would not be an accurate indicator of driver distraction, as it would also entail other exogenous causes.
Mobile phone use affects lane deviation regardless of the use mode (Table 5). The speaker mode involves the highest deviation, being followed by the handheld mode. Experience in using the mobile phone while driving results in improved performance considering lane position. Other contributing parameters include accident involvement (drivers who have been involved in a road accident during the past 3 years exhibit higher lane deviations), driver age, and gender, with drivers older than 44 years old and with women exhibiting higher deviations compared to drivers aged 18-24 and to men, respectively.
Mobile phone is a cause of driver distraction and an emerging cause of road accidents. The use of a mobile phone entails various types of distraction: manual (when its use involves phone handling), visual (when looking at the phone is required), auditory, and cognitive. Most countries adopt the hands-free legislation; that is, using the mobile phone while driving is permitted but handling the device is illegal. Hence, in most countries using the mobile phone under Bluetooth or the speaker mode is allowed. This eliminates manual distraction, while the remainder three types of distraction are still present.
Driver distraction results in the reduction of driver alertness, preventing drivers to react in good time to changes of the road environment whether these involve a pedestrian crossing the road, deceleration of the preceding vehicle, a change of the traffic signal, and so on. Considering the impairment resulting from the manual or cognitive task Consiglio et al.  estimated reaction times for a control scenario, conversation with a passenger, handheld conversation, and hands-free conversation. The resulting mean values (and their corresponding standard deviations) were 0.392 (0.03), 0.453 (0.05), 0.464 (0.41), and 0.465 (0.51). These values can support two hypotheses: “conversation results in distraction” and “the less intrusive the mobile phone use mode is the higher the impairment considering reaction time tends to be.” The findings of this study confirm both hypotheses. Reaction time was found to increase with the use of all three explored mobile phone modes (handheld, hands-free with wired earphones, and speaker). Additionally, the speaker mode was the one exhibiting the highest reaction times.
The reduction of maximum driving speed with the use of mobile phone is a finding noted in all relevant studies. This is yet another consequence of driver distraction but may also be explained through the driver risk compensation mechanism. That is, drivers adjust their driving behaviour so as to reduce the increased risk level of driving, which is a result of mobile phone use. This is achieved through employing lower driving speeds. Driver compensatory behavior has been noted in several relevant studies [22, 40]. Regardless of the cause of speed reduction, this reduction may prove to be critical for the avoidance of an accident or the mitigation of its severity. The results of this study demonstrate a reduction of the maximum driving speed with the use of the handheld mode. Maximum driving speeds also tend to decrease with the use of the hands-free and speaker mode, but this reduction was not found to be statistically significant.
Poorer driving performance, as represented through lane deviation, is another effect of the use of mobile phone while driving. All three explored modes result in higher deviations, with the speaker mode exhibiting the highest effect.
5.2. Limitations and Future Research
Notably the findings of this research should be interpreted also considering the study limitations. Driving behaviour is observed on a driver simulator experiment, which provides an approximation of reality rather than being reality itself. Still, several researchers have shown that the use of driving simulators is a good approximation of reality [41, 42]. Driving in a simulated environment, independently of how realistic this environment is, may make drivers behave somewhat differently compared to being in reality due to a number of reasons (e.g., the driver is observed by external entities and the driver is not exposed to any real danger). However, a strong transfer of the actual driving behaviour in the simulated environment is expected . Within this experiment drivers do not decide whether or not the driving conditions allow for their engagement on the mobile phone conversation, while in real life mobile phone use interactions are self-regulated . Last, the sample size, despite being substantial, cannot fully represent the driving population; and the driving behaviour indicators and driving scenarios used, despite being significant, do not capture all dynamics of actual behaviour and conditions. Concerning sample characteristics, the majority of the participants are young, and this may limit the potential generalisation of findings. In addition, drivers participated on a voluntary basis, which may introduce further bias to the experiment; this however is not expected to alter the relative findings between the different explored scenarios. Concerning behaviour indicators and driving scenarios, further research is targeted towards the investigation of additional driving behaviour indicators (e.g., acceleration, deceleration, and behaviour deviations) and the impact of mobile phone use on driving behaviour at nighttime conditions.
Summarising the effect of mobile phone use on driving behaviour, mobile phone affects driving behaviour irrespective of the use mode, whether it is handheld, with wired earphones, or on speaker mode. Furthermore, the use of the speaker mode entails the highest maximum driving speeds, the highest reaction times, and the highest lane deviation and hence the highest risk compared to using the mobile phone via the handheld mode. Subsequently, the results of this study do not support the adoption of the hands-free mobile phone use legislation. Although there are several other driving behaviour indicators that need to be explored to come to accurate conclusions (such as performing an abrupt manoeuvre to avoid an accident), and probably for some the speaker mode might prove more beneficial than the handheld mode; it is safe to note that the use of the speaker mode is not safe. Thus, legislation on the use of mobile phone while driving should be reconsidered, and relevant parties should consider prohibiting the use of mobile phone while driving under all modes, as is the case in Sweden.
At the same time, driving behaviour was found to be affected less with the increase of familiarity and experience of mobile phone use while driving. For countries where mobile phone use while driving is allowed under specific use modes, efforts to adopt the training curricula and integrate mobile phone use in driver training in an effective manner should be explored. This might be a provocative suggestion, but yet a suggestion that might be beneficial. Putting aside the aforementioned suggestions, in Greece the number of drivers using their mobile phone while driving is particularly high. This is partly due to not only legislation ignorance but also law disobedience. The first requires better campaigning to disseminate the existing legislation. The second requires better campaigning to demonstrate the risks of using the mobile phone while driving and more intense police monitoring and enforcement.
This study stresses the need for further research on the effects of mobile phone use on driving behaviour and road safety and the cooperation of the relevant bodies: academia, policy makers, and enforcement bodies towards ensuring a safer road environment.
The data used in this study can be found via contacting the corresponding author.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Vinci autoroutes fondation, “Despite improved awareness of risks, European drivers are slow to adopt safer behaviour,” Media Release, March 2016.View at: Google Scholar
J. A. Michon, “A critical view of driver behavior models: what do we know, what should we do?” in Human Behavior and Traffic Safety, L. Evans and R. C. Schwing, Eds., Springer, Boston, MA, USA, 1985.View at: Google Scholar
H. Alm and L. Nilsson, “Changes in driver behavior as a function of handsfree mobile phone – a simulator study,” Accident Analysis and Prevention, vol. 4, pp. 441–451, 1993.View at: Google Scholar
A. Benedetto, A. Calvi, and F. D'Amico, “Effects of mobile telephone tasks on driving performance: a driving simulator study,” Advances in Transportation Studies an International Journal Section A, no. 26, pp. 29–44, 2012.View at: Google Scholar
D. Haigney, R. Taylor, and S. Westerman, “Concurrent mobile (cellular) phone use and driving performance: task demand characteristics and compensatory processes,” Transportation Research Part F: Traffic Psychology and Behaviour, vol. 3, no. 3, pp. 113–121, 2000.View at: Publisher Site | Google Scholar
M. M. Haque and S. Washington, “The impact of mobile phone distraction on the braking behavior of young drivers: a hazard-based duration model,” Transportation Research Part C, vol. 50, pp. 13–27, 2015.View at: Google Scholar
A. S. Knapper, M. P. Hagenzieker, and K. A. Brookhuis, “Do in–car devices affect experienced users driving performance?” IATSS Research, vol. 39, pp. 72–78, 2014.View at: Google Scholar
P. Papantoniou, G. Yannis, C. Antoniou, E. Papadimitriou, D. Pavlou, and J. Golias, “Investigating the effects of area type and traffic conditions on distracted driving performance,” Transportation Research Procedia, vol. 14, pp. 3839–3848, 2016.View at: Google Scholar
A. Parkes and V. Hooijmeijer, The influence of the use of mobile phone on driver situation awareness, National Highway Traffic Safety Administration, 2001.
I. A. AlDarrab, Z. A. Khan, and S. I. Ishrat, “An experimental study on the effect of mobile conversation on drivers reaction time in braking response,” Journal of Safety Research, vol. 40, pp. 185–189, 2009.View at: Google Scholar
H. Alm and L. Nilsson, “The effects of a mobile telephone task on driver behavior in a car following situation,” Accident Analysis and Prevention, vol. 27, pp. 707–715, 1995.View at: Google Scholar
P. Papantoniou, C. Antoniou, E. Papadimitriou, D. Pavlou, G. Yannis, and J. Golias, “Is distracted driving performance affected by age? first findings from a driving simulator study,” in Proceedings of the International Interdisciplinary Conference Ageing and Safe Mobility, Bergisch-Gladbach, November 2014.View at: Google Scholar
M. Papadakaki, G. Tzamalouka, C. Gnardelis, T. J. Lajunen, and J. Chliaoutakis, “Driving performance while using a mobile: A simulation study of Greek professional drivers,” Transportation Research Part F: Traffic Psychology and Behaviour, vol. 38, pp. 164–170, 2016.View at: Publisher Site | Google Scholar
C. Saiprasert, S. Supakwong, W. Sangjun, and S. Thajchayapong, “Effects of smartphone usage on driver safety level performance in urban road conditions,” in Proceedings of the 2014 11th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, ECTI-CON 2014, Thailand, May 2014.View at: Google Scholar
D. Lamble, T. Kauranen, M. Laakso, and H. Summala, “Cognitive load and detection thresholds in car following situations: safety implications for using mobile (cellular) telephones while driving,” Accident Analysis & Prevention, vol. 31, no. 6, pp. 617–623, 1999.View at: Publisher Site | Google Scholar
S. Zampino, “The effects of the different uses of mobile phones on drivers: experimental tests on an extra-urban road,” in Proceedings of the AIIT 2nd Congress on Transport Infrastructure and Systems in a Changing World, 2014.View at: Google Scholar
R. A. Tokunaga, A. Shimojo, T. Hagiwara, S. Kagaya, and U. Kenetsu, “Effects of cellular telephone use while driving based on objective and subjective mental workload assessment,” in Proceedings of the 1st International Driving Symposium on Human Factors in Driver Assessment, 2001.View at: Google Scholar
G. Nilsson, “The effects of speed limits on traffic accidents in Sweden,” in Proceedings of the International symposium on the effects of speed limits on traffic crashes and fuel consumption, Dublin, Ireland, 1981.View at: Google Scholar
G. Nilsson, “Traffic safety dimensions and the power model to describe the effect of speed on safety,” Bulletin 221, Lund Institute of Technology, Department of Technology and Society, Traffic Engineering, vol. 221, 2004.View at: Google Scholar