Text messaging while driving: its effects on

driving performance and text messaging

behaviour

Doug Cheung

Project report submitted in part fulfilment of the requirements for the degree
of Master of Science (Human-Computer Interaction with Ergonomics) in the
Faculty of Life Sciences, University College London, 2010.

Note by the University

This project report is submitted as an examination paper. No responsibility

can be held by London University for the accuracy or completeness of the
material therein.
Acknowledgments

Many thanks goes to Duncan Brumby, Justin Grace, and Christian Janssen for
their guidance throughout this project. Thanks also to the participants who
gave up their time to participate in the experiments.

i
 Abstract

Objective: This research investigated the effects of text messag-

ing on simulated driving performance and text messaging behaviour.

Background: In a limited number of studies, text messaging while

driving has been shown to cause impairments in driving performance.

However, drivers’ strategic behaviour to balance the two tasks is still

not fully understood. Method: Sixteen participants used a mounted

touchscreen which simulated a touchscreen mobile phone to read and re-

spond to text messages in both single task (driving) and dual-task (text

messaging while driving) conditions in a fixed-base driving simulator.

Results: Driving performance was significantly impaired when text

messaging while driving, such that lateral deviation increased by 280%

from the single-task to the dual-task condition. In addition, partici-

pants changed their text messaging behaviour by shortening the length

of their replies and typing at a slower speed. Conclusion: Text mes-

saging while driving has a severe negative effect on drivers’ ability to

maintain a central lane position. Attempts to reduce the workload of

the text messaging task by typing shorter messages did not mitigate

the effects of text messaging while driving. Application: The results

provide further evidence for the detrimental effects of text messaging

on driving performance and highlight the effects of multitasking on the

text messaging task. This has potential implications for road safety

measures and future in-car device development.

ii
Contents

1 Introduction 1

2 Literature review 5

2.1 Text messaging while driving . . . . . . . . . . . . . . . . . . . 5

2.2 Manual input while driving . . . . . . . . . . . . . . . . . . . . 7

2.3 Task difficulty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4 Strategy choice . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.5 Current study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Method 12

3.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.2 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.3 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.4 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 Results 16

4.1 Driving performance . . . . . . . . . . . . . . . . . . . . . . . . 16

4.2 Text messaging performance . . . . . . . . . . . . . . . . . . . . 19

5 Discussion 28

5.1 Driving performance . . . . . . . . . . . . . . . . . . . . . . . . 28

5.2 Text messaging performance . . . . . . . . . . . . . . . . . . . . 29

5.3 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.5 Future research . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

6 Conclusion 35

References 36

Appendix A 39

iii
List of Figures

1 Effect of task type and driving speed on RMSE lateral deviation 18

2 Effects of driving speed and text message delay on RMSE lateral

deviation in the dual-task condition . . . . . . . . . . . . . . . . 19

3 Effect of task type and text message delay on message length . 20

4 Effect of task type and text message delay on reaction time . . 21

5 Effect of task type and text message delay on response time . . 22

6 Effect of task type and text message delay on typing speed . . . 23

7 Effect of driving speed and message delay on message length . . 25

8 Effect of driving speed and message delay on reaction time . . . 26

9 Effect of driving speed and message delay on response time . . 26

10 Effect of driving speed and message delay on typing speed . . . 27

iv
1 Introduction

In the UK, the use of any hand-held communications device while driving a
vehicle is illegal, punishable by a monetary fine and 3 penalty points added to
the driver’s licence. Despite severe punishments and government warnings to
discourage drivers from using mobile phones while driving, it is clear that this
activity still takes place. Many surveys have attempted to assess the percentage
of drivers using mobile phones while driving. Although a government survey by
the Department for Transport put the figure at a conservative 1.4% (“Seatbelt
and mobile phone use surveys”, 2010), a recent Yougov poll found that 31%
of drivers admitted to taking a phone call while driving and 22% admitted to
making a call (“One driver in 20”, 2010).

Driving is a safety critical task where the attention, visual and motor skills of
the driver are needed constantly. Mobile device use while driving was deemed
a dangerous enough activity that it was banned in the UK in 2003. However,
the use of hands-free kits remain legal, suggesting that the act of operating
the mobile device is more distracting than the act of holding a conversation.
The law also failed to take into account situational context: it is reasonable
to query whether using a mobile phone during a traffic jam is as dangerous as
using one while travelling at 70 mph on a motorway.

The rise in popularity and ubiquity of mobile phones in recent years has meant
that they are always with us; anyone carrying a mobile phone is reachable wher-
ever and whenever they have a signal. In the modern world, being constantly
connected to family and friends and the almost limitless amount of information
on the World Wide Web is a great convenience afforded by advances in modern
technology. Despite these advantages, mobile devices also have the potential
to distract us when performing safety critical tasks.

Various academic research has been conducted into the use of secondary in-car

1
devices while driving. This research has consistently shown that the use of
such devices can have a dangerous effect on driving performance and increase
the likelihood of having an accident (Drews & Strayer, 2008; Horrey, Wickens,
& Consalus, 2006).

Research on the use of mobile phones while driving has mainly focused on
the act of dialling a telephone number (Brumby, Salvucci, & Howes, 2009;
Salvucci, 2001; Salvucci & Macuga, 2002) (either dialling the number directly
or searching a number from an address book) and having a conversation. There
has been limited research on text messaging specifically and how this activity
affects driving performance; the various factors which affect the success of both
activities simultaneously is still not fully understood (Drews et al, 2009). In
particular, there has been a lack of research focusing on the strategies people
adopt when trying to balance text messaging and driving. This is despite
statistics which suggest that text messaging while driving is almost as common
as having phone conversations while driving: a Yougov poll found that 28%
of drivers confessed to reading a text while driving and 18% said they had
responded to one (“One driver in 20”, 2010).

One reason for this lack of research may be due to the only recent rise in pop-
ularity of text messaging. Mobile phone interfaces have continually evolved to
make input easier and faster: a standard numeric keypad requires the user to
press each number multiple times for one character; a numeric keypad with
predictive text (e.g. T9) software only requires one key press per letter; phones
with full qwerty keyboards allow a separate key press per letter as on a stan-
dard computer keyboard; phones with touchscreen qwerty keyboards can be
language independent and change on-screen keys according to context, for ex-
ample showing a “.com” button when typing an email.

The act of text messaging shares common ground with other in-car secondary
tasks, but is also different in some ways. To demonstrate this, we shall examine

2
the act of dialling a telephone number. The driver has the phone number
memorised and has to enter the phone number using the keypad with one
hand, while steering with the other hand. Motor skills are divided between
the two tasks, preventing the driver from performing many driving-related
tasks such as changing gears while operating the device. Entering the phone
number also requires the driver’s vision and attention; when looking at the
phone, pressing a key and looking at the display, the driver cannot see what
is happening on the road and cannot perceive and react to hazards such as
a pedestrian crossing the road or the car in front performing an emergency
brake. However, the driver can choose which strategy to adopt when dialling
while driving: for example, dialling one number at a time and interleaving
attention back to driving frequently, or dialling all numbers at once and not
interleaving attention at all, or any other variation of strategy. Research has
shown that cognitive constraints such as the representational structure of the
phone number can have an effect on the strategy adopted (Brumby, Salvucci,
& Howes, 2009).

The issues described so far are common to both dialling a phone number and
text messaging, as well as operating a satellite navigation system or in-car
entertainment system. However, text messaging differs to dialling in that there
is a further element of choice, in addition to choice of interleaving strategy:
the driver can choose what to write in the text message. Unlike with phone
number dialing, the driver can choose how to long to make the message, how
much information to include, and how to represent the information. When
responding to a message, the driver can also choose when to start typing a
response. Various factors may affect driving performance and text messaging
choice and performance in this scenario. One of these factors is the difficulty
of the driving task: for example, do drivers send shorter messages and react
slower when driving at high speed compared to low speed? Another factor is
frequency of text messaging: for example, do drivers react quicker if a new

3
incoming message is received immediately after a response has just been sent?
How does driving performance change as a result of these factors? A better
understanding of questions such as these may help the design of safer systems.

This dissertation investigates how task difficulty and task frequency affect driv-
ing performance, text messaging performance, and strategy choice. Section 2
presents a review of relevant literature in the use of in-car devices while driv-
ing, with a particular focus on text messaging. Section 3 describes the method
of a controlled laboratory experiment that investigated driver behaviour when
responding to text messages while driving in a fixed-base driving simulator.
Section 4 presents a summary of the results of the experiment and section 5
discusses the findings in relation to current literature, as well as discussing
implications for the design of mobile and in-car devices. Limitations of the
study are acknowledged and future areas of research are identified.

4
2 Literature review

2.1 Text messaging while driving

In the limited number of empirical studies which have focused exclusively on

the effects of text messaging on driving, all have shown text messaging to have
adverse effects on various metrics of driving performance.

Drews, Yazdani, Godfrey, Cooper, & Strayer (2009) conducted a study into the
effects of text messaging on participants’ performance in a high-fidelity driving
simulator. In the experiment, pairs of participants exchanged text messages
with one another with the goal of arranging an evening activity, while one par-
ticipant drove along a freeway road behind a leading pace car. The leading car
braked intermittently to assess the impact of text messaging on participant’s
ability to react to road events. Results showed a significant decrease in driving
performance across all measured metrics: under the dual-task condition com-
pared to the baseline single-task driving condition, participants had slower
brake onset times, more lane departures, reduced lane maintenance overall,
and notably a six-fold increase in crash rate. Interestingly, participants in-
creased their following distance in the dual-task condition; this may have been
a conscious or unconscious attempt to create a safety buffer with the leading
car to reduce crash likelihood. However, this strategy proved inadequate, as
evidenced by the high crash likelihood.

A study focusing specifically on young drivers found similar results. Hosking,

Young, & Regan (2009) investigated the effects of text messaging on the driv-
ing performance of young novice drivers. Young drivers were chosen as they
were found to be more willing to text than experienced drivers, while not hav-
ing the experience necessary to drive with limited attentional resources. Young
drivers were also considered more vulnerable to distraction: it was predicted
that more resources would be allocated to driving, leaving fewer resources for

5
secondary tasks and thus impacting driving performance when tasks needed to
be balanced. The experiment involved participants text messaging while driv-
ing in a driving simulator, with 8 test events to evaluate participants’ driving
performance. Results found a significant increase in the number of missed lane
changes in the car-following event; an increase in the number of lane excursions
overall; and an increase in drivers’ mean time headway (distance from leading
car) and time headway variability overall. This increase in mean time head-
way mirrors the behaviour seen in the Drews et al. study, where participants
created a safety buffer behind the leading car. In addition, eye movement
analyses (recorded using eye-tracking software) found a significant increase in
the proportion of time looking inside the vehicle, an increase in the number of
in-vehicle glances, and an increase in the length of time of in-vehicle glances.

In an earlier small-scale study featuring 10 participants, Kircher et al. (2004)

conducted an experiment focusing only on receiving text messages, with par-
ticipants responding to the text message verbally. Results found a significantly
increased braking time for one of the four simulated hazard conditions, with
no other significant effects found. However, this could be due to the small
sample size and the use of experienced drivers, whose driving skill were at a
level where a secondary task did not noticeably impact driving performance.

Clearly, the driving conditions used by Drews et al. in their experiment were
much more extreme than typical everyday driving. Participants drove in two
scenarios: in each scenario, the leading pace car braked at 42 randomly selected
intervals, providing a total of 84 opportunities for each participant to crash.
Coupled with the high speeds typically encountered on a freeway, this highly
dangerous condition of a frequently braking lead car could be part of the reason
for the poor driving performance found when participants were text messaging
while driving.

Hosking et al.’s study deliberately focused on young inexperienced drivers who

6
were found to be more willing to text message while driving. Their prediction
that young novice drivers would lack the skill necessary to perform a sec-
ondary task without adverse effects on driving performance was found to be
correct. However, surprisingly, analyses of participants’ driving performance
in response to three hazard conditions (traffic light, pedestrian crossing the
road, and a right-turning car) found no significant effects on braking reac-
tion times when text messaging while driving. The authors note limitations
in methodology as possible reasons for this, including participants learning to
expect hazards and the experience the participants had with text messaging.
It should be noted that the text messaging task was very simple: participants
only had to respond to text messages with one-word replies. This may also
have been a factor in participants’ success in responding to road hazards.

2.2 Manual input while driving

Despite the lack of research specifically in the area of text messaging while
driving, we can also draw on studies which feature other manual input tasks
while driving.

Tsimhoni, Smith, & Green (2004) compared input methods for address en-
try on an in-car navigation system. One of the input methods assessed was
a touchscreen keyboard, very similar to a mobile phone with an on-screen
touchscreen keyboard. A driving simulator study found a serious degradation
of vehicle control when using the touchscreen for manual input while driving,
with increases in lateral deviation and lane departures from the single-task
driving condition. Furthermore, the address entry task also took significantly
longer to complete. However, task differences may mean the results here can-
not be directly applied to text messaging while driving. In the address entry
task, participants retrieved addresses from a separate memo display before
typing them using the touchscreen. In a text messaging task, this is similar to

7
reading an incoming message, but text messaging also requires more cognitive
effort to think of a response and hold it in working memory before typing it
out. Therefore the effects of text messaging may be even greater than those of
address entry while driving.

Dialling a telephone number is also a task similar to text messaging. Re-

search has consistently shown the detrimental effects of dialling on driving
performance (Brumby, Salvucci, & Howes, 2009; Salvucci, 2001; Salvucci &
Macuga, 2002). See Collet, Guillot, & Petit (2010) for a comprehensive review
of epidemiological, psychological, behavioural and physiological studies on the
impact of mobile phone use on driving performance. Regan, Young, Lee, &
Gordon (2008) also performed a task analysis of the steps involved in talking
on a mobile phone, finding that the act of manually dialling a phone number
has the most detrimental impact on driving performance. These findings sug-
gest that text messaging, which features prolonged periods of manual input of
complex information, may significantly impair driving performance.

2.3 Task difficulty

Research has shown that the demands of the driving task and the demands of
the secondary task have an effect on driving performance when multi-tasking
(Horrey & Wickens, 2004; Lee, Regan, & Young, 2008). In the Drews et al.
study, the high frequency with which participants had to brake may have had
an effect on the resulting findings: high task difficulty may have contributed to
the high crash rate found in the dual-task condition. The simplicity of the text
messaging task, only requiring participants to give one-word replies, may have
been a factor in Hosking et al.’s study: the low difficulty of the text messaging
task may have contributed to participants’ successful driving performance in
the dual-task condition. Therefore, task difficulty - both of the driving task
and the text messaging task - may be an important factor in determining the

8
extent to which text messaging while driving affects driving performance.

Driving speed was used to manipulate driving difficulty in an experiment by

Brumby et al. (2009) to investigate strategy adaptation in a dual-task setting
involving dialling a telephone number while driving. Results found a significant
effect of speed on driver’s ability to maintain a central lane position in both the
single-task driving and the dual-task dialling while driving conditions. Driving
performance was significantly worse when driving at a high speed (55 mph)
compared to a low speed (35 mph). The steps involved in text messaging is
very similar to dialling in that both tasks require manual input from the driver
and therefore a sharing of visual attention and motor skills. It is therefore
reasonable to assume that different driving speeds may have an impact on
driving performance when text messaging while driving.

The experiments conducted by Drews et al., Hosking et al., and Kircher et

al. all did not place emphasis on the text messaging task: Drews et al. had
pairs of participants arranging an evening event via text message, but did
not control this task nor report associated measures; Hosking et al. only
required participants to reply to text messages using one-word answers; and
Kircher et al. did not require participants to reply via text message at all,
but simply retrieve the incoming message and respond verbally. Looking at
text messaging performance in addition to driving performance may allow us
to better understand drivers’ behaviour.

2.4 Strategy choice

One area where analysing text messaging performance could be beneficial is

strategy selection: how do drivers attempt to balance the two task objectives
when text messaging while driving? How does strategy selection change in
response to changing the demands of the driving task and the text messaging

9
task? An indication to these answers can be found in Horrey & Lesch’s (2009)
study which examined whether drivers would strategically select the best op-
portunities to carry out distracting in-vehicle tasks, given prior knowledge of
upcoming road demands. In the experiment, participants drove around a test-
track and were instructed to perform an in-vehicle task such as reading a text
message; participants had free choice when to carry out the task and knew
which sections of the test-track were more difficult to navigate. Results found
no strategic adaptation: participants carried out the task regardless of the road
demands and did not postpone tasks until they encountered an easier section
of the course. This frequently led to driving errors: participants made an error
in 19% of the trials, providing evidence that in-vehicle tasks negatively affect
driving performance even when drivers have free choice of when to perform
them.

Applying these results to specifically to text messaging while driving, this may
suggest that regardless of the difficulty of the driving task, drivers would not
alter their text messaging behaviour, with subsequent detrimental effects on
driving performance. On the other hand, if drivers do strategically adapt to
the demands of the driving task, then in difficult conditions such as driving at
a high speed, we would expect reaction times to text messages to be longer,
messages to be shorter, and typing speed to be slower, as more focus is given
to the driving task in order to maintain lane position.

2.5 Current study

The first goal of this study is to provide further empirical evidence in support of
or against the claim that text messaging while driving has a detrimental effect
on driving performance. To simplify the experiment, lateral deviation from
the centre of the lane will be the only measure of driving performance used.
Unlike braking reaction times, lateral deviation does not depend on external

10
events such as road hazards which are difficult to simulate realistically.

The second goal of this dissertation is to explore factors which may have an
effect on the extent to which text messaging affects driving performance. Driv-
ing speed will be manipulated in order to change the difficulty of the driving
task. Delay times between incoming text messages will be changed in order to
vary the frequency of the text messaging task.

The third goal of this dissertation is to investigate strategy selection. To facil-

itate this, attention will be paid to the performance of the text messaging task
in addition to the performance of the driving task. It is hoped this approach
will capture any performance trade-offs which may or may not differ between
the task types, driving speeds, and text message delay times.

11
3 Method

3.1 Participants

16 postgraduate students from UCL took part in the experiment. All partici-
pants held a valid UK driving license with a mean driving experience of 6.46
years (standard deviation = 3.33). All participants were screened for normal
or corrected to normal visual acuity. Ages of participants ranged from 22 to 37
years, with a mean of 27.56 years (standard deviation = 4.1). 10 participants
were male. Participants were not paid nor did they receive course credit for
taking part in the experiment. Participants were warned not to take part if
they suffered from motion sickness.

3.2 Equipment

A low-fidelity fixed-base driving simulator was used, which consisted of: an

Apple Mac Mini; the driving simulator software; a Logitech driving wheel; and
a 30 inch LCD screen as the main display, at a resolution of 1280 x 800. A
mobile phone simulator was used, which consisted of: a 7 inch LCD touch-
screen as the secondary display and input device, at a resolution of 800 x 600;
and text messaging software created using Java and modelled on the Apple
iPhone’s text messaging system. The secondary display was positioned to the
left of the main display, approximately where a real mobile phone would be
mounted to a vehicle’s centre console.

To make the text messaging task as realistic as possible, questions sent to par-
ticipants as text messages were created from Thurlow & Brown’s (2003) anal-
ysis of communication orientations and themes in text messages. The analysis
identified key categories of text messages such as informational-practical (e.g.
asking where a place is) and practical/social-arrangement (e.g. arranging the

12
time to meet for an activity). Question types were balanced across all condi-
tions in the experiment in order to remove the likelihood that a set of questions
in one particular condition would elicit shorter or longer responses, or be more
or less difficult to process and respond to.

3.3 Design

The single-task trials were used to obtain baseline measurements of driving

and text messaging performance.

In the dual-task trials, a 2x2 within-subjects design was used, with all par-
ticipants carrying out all conditions of the experiment. The conditions were
systematically counter-balanced to account for order effects; 4 participants
carried out the experiment for each order of conditions. The independent vari-
ables were driving speed (slow: 30mph and fast: 70mph) and the time delay
between text messages (short: 3 seconds and long: 15 seconds).

For the single-task driving and dual-task conditions, the dependent variable
was: root-mean-squared error (RMSE) lateral deviation (metres).

For the single-task text messaging and dual-task conditions, the dependent
variables were: text message verbosity (number of characters, reaction time
(time from the incoming message being displayed to the participant’s first key
press), response time (time from the participant’s first key press to their last
key press), and typing speed (number of characters per second).

3.4 Procedure

Participants were first given an information sheet describing the purpose of the
experiment and asked to sign a consent form. They were given the opportunity
to ask questions before the experiment started.

13
The practise trials allowed participants to gain familiarity with the driving
simulator and the mobile phone simulator. Participants practised for one trial
each of driving at slow speed, driving at fast speed, and text messaging.

In the driving single-task, participants had to steer an auto-accelerating car,

travelling at a constant speed, down a straight road (the middle lane of a
three lane motorway). Traffic cones were displayed at the lane boundaries to
discourage lateral movement. Steering to maintain a central lane position was
the only driving task; no pedals were used.

In the text messaging single-task, participants received text messages on the

touch-screen secondary display, which simulated a touch-screen mobile phone.
A received text message was signified by a beeping sound to alert the partici-
pant, with the text message instantly appearing on the display. The secondary
display featured a on-screen ’QWERTY’ keyboard which was used by partic-
ipants to type in their responses. A ’send’ button was used to confirm the
sending of a response. After a response was sent, and after a delay period (3
seconds or 15 seconds depending on delay condition), participants were sent
another message to respond to. There were a total of 3 incoming messages
per trial. Participants were instructed to respond to all 3 messages per trial,
but were not given any instructions on when to respond, nor what to type in
response.

In the single-task trials, participants performed the driving or texting tasks

in isolation: 2 trials each of driving at slow speed, driving at fast speed, text
messaging with a short delay, and text messaging with a long delay. In each
driving trial, participants drove for 20 seconds, followed by feedback on their
RMSE lateral deviation. In each text messaging trial, participants responded
to 3 text messages, followed by no feedback.

In the dual-task conditions, participants performed the driving and texting task
simultaneously: driving and texting at slow speed and high speed, with short

14
and long delays between text messages. Each trial was followed by feedback on
their RMSE lateral deviation. In total, each participant completed 8 dual-task
trials: 2 trials x 2 driving speeds x 2 text messages delays. In the dual-task
trials, each participant responded to a total of 24 text messages: 8 dual-task
trials x 3 messages per trial.

15
4 Results

The dependent measures of interest varied depending on the trial type. For
the driving single-task: the RMSE lateral deviation of the participant’s vehicle
from the centre of the lane was the only dependent variable; for the text
messaging single-task: the message length, reaction time, response time, and
typing speed of the participant were the dependent variables. For the dual-task
trials, all dependent variables were of importance.

Data from one participant was excluded from data analysis due to having a
mean driving score of more than two standard deviations above the mean for
all participants.

From a total of 180 (4 single-task and 8 dual-task trials per participant; 15

participants) trials involving the touchscreen and the custom-made text mes-
saging software, data from only one trial was excluded from analysis; this was
due to a touchscreen malfunction during a text messaging single-task trial.
The trial had to be interrupted to reset the touchscreen, leading to artificially
inflated reaction and response times, hence why the data was excluded.

The next sections describe the results of a detailed statistical analysis of

the data obtained from the experiment. Unless otherwise stated, a two-way
repeated-measures ANOVA with variables of driving speed (fast; slow) and
text message delay (long; short) was used for statistical analyses. Both vari-
ables were analysed as within-subject variables. An alpha level of .05 was used
throughout.

4.1 Driving performance

The only dependent variable on which driving performance was measured was
the RMSE lateral deviation across the trial. The driving simulator logged the

16
lateral deviation of the participants’ vehicle from the centre of the lane at a
rate of 200Hz. The RMSE of these cumulative lateral deviation samples was
then calculated over the the duration of the trial.

The first analysis compared driving performance between the two task types
and driving speeds. The second analysis compared driving performance be-
tween the dual-task conditions of driving speed and text message delay.

4.1.1 Effect of multi-tasking on driving performance

Analysis compared driving performance between the single-task driving con-

dition and dual-task text messaging while driving condition under differing
driving speeds. Figure 1 shows the effects of task type (single-task; dual-task)
and driving speed (slow; fast) on participants’ RMSE lateral deviation. Lat-
eral deviation was significantly lower in the single-task condition (M = .320m,
SD = .095m) compared to the dual-task condition (M = .906m, SD = .370m),
F (1,14) = 50.8, p < .001. This shows that participants’ driving performance
was better when only driving, compared to text-messaging while driving.

Analysis also showed that there was no main effect of driving speed on lateral
deviation (p = .353). There was also no significant interaction between driving
speed and task type, p = .342.

17
 Figure 1: Effect of task type and driving speed on RMSE lateral deviation

4.1.2 Effects of driving speed and text message delay on driving

performance in the dual-task condition

Figure 2 shows the effect of driving speed and text message delay on driving
performance (RMSE lateral deviation) in the dual-task condition. There was
a significant main effect of text message delay (long; short) on driving perfor-
mance (F (1,14) = 11.3, p = .005); participants’ deviation was lower in the long
delay condition (M = .820m, SD = .368) compared to the short delay condition
(M = .992, SD = .434). This shows that participants’ driving performance
was better when there were long delays between incoming text messages, com-
pared to short delays between incoming text messages. However, there was
no significant main effect of driving speed on driving performance, p = .320.
Furthermore, there was no significant interaction between driving speed and
text message delay, p = .317.

18
Figure 2: Effects of driving speed and text message delay on RMSE lateral
deviation in the dual-task condition

4.2 Text messaging performance

Text messaging performance was measured using multiple dependent variables:

message length, reaction time, response time, and typing speed. A full descrip-
tion of these variables can be found in the Method section in section 3.

The first analysis compared text messaging performance between the two task
types and text message delay times. The second analysis compared text mes-
saging performance between the dual-task conditions of driving speed and text
message delay.

4.2.1 Effect of multi-tasking on text messaging performance

Analysis compared text messaging performance between the single-task text

messaging condition and dual-task text messaging while driving condition un-

19
der differing text message delay times. Analysis showed significant main effects
of task type (single; dual) on the dependent variables of message length, reac-
tion time, and typing speed.

Figure 3 shows the mean message lengths for the single-task and dual-task
conditions under the short and long text message delay times. There was a
significant main effect of task type on message length: message length was
statistically significantly lower in the dual-task condition (M = 15.9, SD =
7.23) compared to the single-task condition (M = 28.0, SD = 12.1), F (1,14)
= 32.8, p < .001. This shows that participants typed shorter messages when
text messaging while driving, compared to only text messaging. There was no
significant main effect of driving speed on text message length, p = .302. There
was a significant interaction between task type and driving speed, F (1,14) =
5.41, p = .036.

Figure 3: Effect of task type and text message delay on message length

Figure 4 shows the mean reaction times for the single-task and dual-task con-

20
ditions under the short and long text message delay times. There was a signif-
icant main effect of task type on reaction time: reaction time was statistically
significantly higher in the dual-task condition (M = 8.72s, SD = 2.86s) com-
pared to the single-task condition (M = 4.49s, SD = 1.73s), F (1,14) = 49.7,
p < .001. This shows that participants took longer to start typing a response
to an incoming message when text messaging while driving, compared to only
text messaging.

There was also a significant main effect of text message delay time on reaction
time: reaction time was lower when there were short delays between incoming
text messages (M = 7.04s, SD = 3.37s) compared to long delays (M = 7.58s,
SD = 3.41s), F (1,14) = 7.49, p = .016. This shows that participants reacted
quicker to incoming messages when they were received quicker following an
outgoing message. There was no significant interaction between task type and
text message delay time, p = .612.

Figure 4: Effect of task type and text message delay on reaction time

21
Figure 5 shows the mean response times for the single-task and dual-task
conditions under the short and long text message delay times. There was no
significant main effect of task type on response time, p = .878. Response
times for the single-task text messaging condition (M = 28.9s, SD = 10.6s)
were approximately the same as the dual-task text messaging while driving
condition (M = 29.4s, SD = 13.4s). This shows that participants spent the
same amount of time typing in responses regardless of whether multi-tasking
or single-tasking.

There was also no significant main effect of text message delay time on response
time, p = .858. Participants spent the same amount of time typing in responses
regardless of whether multi-tasking or single-tasking. There was a significant
interaction between task type and text message delay time, F (1,14) = 4.93, p
= .043.

Figure 5: Effect of task type and text message delay on response time

Figure 6 shows the mean typing speeds for the single-task and dual-task con-

22
ditions under the short and long text message delay times. There was a signif-
icant main effect of task type on typing speed: typing speed was statistically
significantly lower in the dual-task condition (M = .674, SD = .133) compared
to the single-task condition (M = 1.19, SD = .239), F (1,14) = 82.1, p < .001.
This shows that participants typed slower when text messaging while driv-
ing, compared to only text messaging. There was no significant main effect of
driving speed on typing speed, p = .383. There was no significant interaction
between task type and driving speed, p = .718.

Figure 6: Effect of task type and text message delay on typing speed

In summary, participants took longer to react to text messages, typed shorter

responses, and typed slower in the dual-task condition compared to the single-
task condition. However, there was no difference in response time, the time it
took to type in and send a response.

23
4.2.2 Effects of driving speed and text message delay on text mes-
saging performance in the dual-task condition

Analysis compared the effects of driving speed (slow; fast) and text message
delay (short; long) in the dual-task condition on the dependent variables used
to measure text messaging performance. In the analysis of typing speed, there
was a trend interaction effect between driving speed and text messaging delay,
F (1,14) = 3.92, p = .068. There were no other statistically significant effects
found.

Figure 7 shows the effects of driving speed and text message delay on the
mean message length. There was no significant main effect of driving speed
on message length (p = .124) nor text message delay on message length (p
= .698). There was no significant interaction between driving speed and text
message delay, p = .293.

Figure 8 shows the effects of driving speed and text message delay on the mean
reaction time. There was no significant main effect of driving speed on reaction
time (p = .184) nor text message delay on reaction time (p = .290). There
was no significant interaction between driving speed and text message delay,
p = .255.

Figure 9 shows the effects of driving speed and text message delay on the
mean response time. There was no significant main effect of driving speed on
response time (p = .630) nor text message delay on response time (p = .171).
There was no significant interaction between driving speed and text message
delay, p = .446.

Figure 10 shows the effects of driving speed and text message delay on the
mean typing speed. There was no significant main effect of driving speed on
typing speed (p = .141) nor text message delay on typing speed (p = .558).

24
These results suggest that the independent variables of driving speed and text
message delay did not significantly affect participants’ text messaging perfor-
mance in the dual-task condition.

Figure 7: Effect of driving speed and message delay on message length

25
Figure 8: Effect of driving speed and message delay on reaction time

Figure 9: Effect of driving speed and message delay on response time

26
Figure 10: Effect of driving speed and message delay on typing speed

27
5 Discussion

5.1 Driving performance

The experiment found participants’ lateral deviation to be significantly worse

when text messaging while driving compared to when driving only. There was
a 0.586m increase in lateral deviation from 0.320m in the single-task condition
to 0.906m in the dual-task condition; a 280% increase. This supports previous
research in this area which has found text messaging to be detrimental to
driving performance (Kircher et al., 2004; Drews et al., 2009; Hosking et al.,
2009); these studies have highlighted various reasons for this finding, with each
having varying levels of applicability in this study.

It is likely that the visual demands of text messaging was a highly important
factor in the impairment of driving performance. Due to the position of the
touchscreen, participants interleaved visual attention between the touchscreen
and the road. Studies have shown the costs of display separation to be greater
for head-down displays compared to adjacent displays (Horrey & Wickens,
2004). This suggests driving performance may have improved if participants
were allowed to share visual attention between the touchscreen and the road by
holding the touchscreen at eye level while driving. The amount of time spent
not looking at the road while driving contributes to detriments in driving
performance and increases in crash risk (Lansdown, 2001).

Compared to other manual input tasks such as dialling a phone number, text
messaging requires additional cognitive resources to process the message and
understand its meaning. It is likely the cognitive demands of the text messag-
ing task was a factor in the impairment of driving performance.

As the driving task was simple, requiring participants to only steer the vehicle
to correct lane position along a straight road, the sharing of motor skills was

28
unlikely to be an important factor in the impairment of driving performance.
The steering wheel could easily be adjusted with just one hand, leaving the
other hand free for operating the touchscreen.

The difficulty of the driving task, as modified by the driving speed, did not have
an effect on lateral deviation. This was seen in both the single-task and dual-
task conditions, and in the combined data. This supports Horrey & Lesch’s
(2009) findings that drivers do not strategically adapt dual-task strategy ac-
cording to the demands of the driving task, with subsequent negative effects
on driving performance. Further evidence can found in the results obtained for
text messaging performance which were the same both driving speeds. This
will be discussed in the following section on text messaging performance.

Looking at the data from the dual-task condition only, the text message de-
lay time had an effect on lateral deviation: short delays between incoming
messages led to worse driving performance. This shows that the frequency
of interruption from the text messaging task affected participants’ ability to
maintain a central lane position. This is likely a result of participants spend-
ing less time correcting the vehicle’s position on the road over the course of
the trial due to more frequent incoming text messages. However, participants
could have chosen to spend the same amount of time on the driving task in
both delay conditions: in the shorter delay condition, the text message could
have simply been ignored until the participant felt the vehicle was heading in a
stable direction. In fact, all text message performance variables were the same
in both delay conditions.

5.2 Text messaging performance

An aspect of text messaging while driving which hasn’t been studied is per-
formance of the text messaging task. Comparing text messaging performance

29
between the single-task and dual-task conditions, there is a clear detrimen-
tal effect of multi-tasking in the dual-task condition. The experiment found
participants to type significantly shorter messages when text messaging while
driving compared to when text messaging only: there was a 12.1 character
decrease in message length from 28.0 characters in the single-task condition to
15.9 characters in the dual-task condition. Reaction time was also significantly
longer: there was a 4.23s increase in reaction time from 4.49s in the single-task
condition to 8.72s in the dual-task condition. Typing speed was also signifi-
cantly lower: there was a 0.516 characters per second decrease in typing speed
from 1.19 in the single-task condition to 0.674 characters per second in the
dual-task condition. Response time, however, was the same across both condi-
tions. It seems likely that message length counterbalanced with typing speed
to produce similar response times: typing shorter messages at a slower speed
took approximately the same amount of time as typing longer messages at a
faster speed.

Participants were given free choice of the content and length of their replies,
so the effect of producing shorter messages when text messaging while driving
can be seen as a strategy adaptation, whether conscious or subconscious, to
reduce the workload of the text messaging task in the dual-task condition.
Despite this strategy, driving performance was still severely impaired in the
dual-task condition. Longer reaction times are likely to be a result of partici-
pants wanting to correct the heading of the car before switching attention to
the touchscreen to read and process the text message. Reductions in typing
speed are likely to be the result of interleaving attention between the driving
task and the text messaging task. If instead participants simply typed the
whole text message before returning attention back to the driving task, we
would expect typing speeds to be the same in both task conditions, with even
more severe impairments to driving performance in the dual-task condition.

Overall, text message delay time also affected reaction time: participants re-

30
acted to text messages quicker when there was a shorter delay between incom-
ing messages. Shorter text message delay times would have allowed partic-
ipants to more easily predict when an incoming message would be received;
longer text message delay times would have meant attention being fully fo-
cused back onto the driving task, with incoming text messages being more of
a surprise. Delay time had no other effect on text messaging performance.

Driving speed had no effect on any variables used to measure text messaging
performance. As mentioned earlier, this supports research which has found
drivers not to strategically adapt to difficulties in the driving task in dual-task
scenarios. If participants did strategically adapt to the fast driving speed, we
would expect to see more focus given to the driving task, with subsequently
longer reaction times, shorter messages, and slower typing speeds.

5.3 Implications

This study has provided further evidence of the detrimental effects of text
messaging on driving performance, supporting the ban on the use of hand-held
communications devices while driving. Crucially, the level of impairment was
equal regardless of the driving speed, suggesting that it is equally as dangerous
to text message at slow speeds and high speeds. One of the reasons for this
finding is that drivers did not alter their approach to the text messaging task
depending on the difficulty of the driving task. However, this may not be true
in real life, where there are real implications for lateral deviation.

The visual demand of the text messaging task was likely the most important
factor in influencing driving performance. Simply using a hands-free device
may not be sufficient to mitigate the risks of in-car device use; an eyes-free
device is also important. Indeed, studies have shown driving performance to
be the same for drivers using handheld and hands-free mobile phones (Strayer
& Johnston, 2001; McEvoy et al, 2006).

31
The change in text messaging performance shows strategic adaptation sensitive
to the text messaging task. A direct design implication of this would be to
design mobile phones which explicitly allow the user to reduce the number of
key presses, since both T9 predictive text entry and full keyboard entry require
the user to carry out one key press per character. Speech based systems are an
alternative that is actively being researched, although these systems may place
additional cognitive demands on the driver (Jamson, Westerman, Hockey, &
Carsten, 2004; Lee, Caven, Haake, & Brown, 2001). If text messaging while
driving cannot be completely avoided, it is at least helpful that drivers be
educated about the costs and risks associated with interacting with electronic
devices while driving. This may encourage a stronger focus on driving and a
reduced willingness to reply to text messages until it is safe to do so.

5.4 Limitations

Data from one participant was excluded from analysis because of an average
deviation score of more than two standard deviations above the mean across
all conditions. This suggests the participant did not learn the driving task
sufficiently, or prioritised the text messaging task above the driving task. Due
to time limitations and following an analysis of the driving task, it was deemed
that two driving practise trials would be sufficient for participants to learn the
driving task. Insufficient training may have been the cause of this outlier, and
may have led to poorer driving performance overall and an increase in variation
of driving performance. Participants were not explicitly told to prioritise the
driving task above the text messaging task, but since only driving performance
was reported at the end of each trial, it was clear that the driving task should
be given more focus to improve lateral deviation.

The small sample size may also have affected the findings from statistical
analysis; the low of statistical power of the analysis of the limited data set

32
may have failed to find significant effects, leading to type II errors. However,
this aspect of the experiment was unavoidable due to resource constraints
restricting the number of participants available and the time available in the
laboratory.

The low realism of the driving simulator and text messaging system could have
a detrimental impact on the ecological validity of the findings. The driving task
was artificially simple and may have not reflected the everyday demands of real
life driving, causing multi-tasking to be easier. Conversely, participants fre-
quently reported the steering to be unrealistically sensitive, with small steering
adjustments leading to large on-screen movements, increasing the difficulty of
the driving task.

Similarly, the text messaging system also had issues with difficulty. Partici-
pants frequently complained about the unresponsiveness of the touchscreen,
leading to frequent errors and frustration. The touchscreen required relatively
long and deliberate presses to register a key press, compared with the respon-
siveness and accuracy of modern touchscreen phones. The low ease of use of the
text messaging system may have increased the difficulty of the text messaging
task, leading to modified task performance. In addition, despite a fitting trial
in a pilot study, it seems the location and position of the touchscreen was not
optimal. The touchscreen was placed approximately where an in-car phone
holder would be mounted near to the centre console. However, participants
reported that in real life, they would prefer to hold the phone themselves in
order to have better peripheral vision of the road. Participants who were not
UK citizens also expressed a preference for the touchscreen to be on the right,
as they were used to driving left-hand-drive cars.

Aside from technical issues, the study is of course a simulator study which
also carries its own limitations. The outcome of the driving task and the
text messaging task had no real world effect on the participants, so they were

33
relied upon to perform as realistically as possible. Regardless of the driving
performance, there was no physical risk to participants, so it is fair to assume
there was more attention given to the text messaging task than in real life
where accidents are a very real risk. On the other hand, drivers may have
more motivation to engage in text messaging in real life in certain situations.
For example, a person driving to an event may receive a text message from their
friend telling them the event has changed locations. The new location of the
event is highly important information as the driver may need to change driving
direction before it is inconvenient to do so, for example exiting a motorway at
the next junction. This may give the driver more incentive to focus on text
messaging at the expense of driving.

5.5 Future research

One main improvement to this study would be to increase the realism of the
driving task. This includes improving the realism of the driving simulator
and increasing the variety and difficulty of the driving task itself. Driving is
not a simple steering task, and simulator studies should aim to replicate the
demands encountered in everyday driving in order to better assess the impacts
of engaging in secondary tasks. A more complex driving task would also allow
the collection of different types of data to more accurately measure driving
performance, such as braking reaction times and lane changing accuracy.

An interesting approach used by Hosking et al. (2009) is the use of eye-tracking

software. This allowed for a deeper analysis of driver behaviour by analysing
eye movements during the experiment. This would have been beneficial in this
study as it would have enabled an analysis of driver interleaving behaviour
between the text messaging task and the driving task. Future research could
use eye-tracking software to develop a better understanding of visual attention
switching when text messaging and driving.

34
6 Conclusion

A limited number of previous studies have shown the detrimental effects of

text messaging on driving performance. This study provided further evidence
to these claims, and also investigated the effect of multitasking on text mes-
saging behaviour. Other factors which modified the difficulty of the driving
task and frequency of interruption were also explored. Drivers adapted their
text messaging behaviour to cope with the demands of balancing the text mes-
saging and driving tasks, however this was not successful in eliminating the
impairments in driving performance. The difficulty of the driving task had no
effect on any performance criteria, both of the the driving task and the text
messaging task. This suggests an insensitivity to the difficulty of the driving
task, as seen by Horrey & Lesch (2009). Given the dangers of interacting with
in-car devices while driving, particularly those which require frequent visual
attention and have the potential to abruptly distract the driver, it is recom-
mended that measures be implemented to reduce these risks to improve driving
safety.

35
References

Atchley, P., & Dressel, J. (2004). Conversation Limits the Functional Field of
View. Human Factors, 46(4), 664-673. doi:10.1518/hfes.46.4.664.56808

Brumby, D. P., Salvucci, D. D., & Howes, A. (2009). Focus on driving: how
cognitive constraints shape the adaptation of strategy when dialing while driv-
ing. In Proceedings of the 27th international conference on Human factors in
computing systems (pp. 1629-1638). Boston, MA, USA: ACM.

Collet, C., Guillot, A., & Petit, C. (2010). Phoning while driving: a review
of epidemiological, psychological, behavioural and physiological studies. Er-
gonomics, 53(5), 589-601.

Drews, F. A., & Strayer, D. L. (2008). Cellular phones and driver distraction.
In M. A. Regan, J. D. Lee & K. L. Young (Eds.), Driver distraction: Theory,
effects, and mitigation (pp. 169-190). Boca Raton, FL: CRC.

Drews, F. A., Yazdani, H., Godfrey, C. N., Cooper, J. M., & Strayer, D. L.
(2009). Text Messaging During Simulated Driving. Human Factors, 51(5),
762-770.

Horrey, W. J., & Lesch, M. F. (2009). Driver-initiated distractions: Exam-

ining strategic adaptation for in-vehicle task initiation. Accident Analysis &
Prevention, 41(1), 115-122.

Horrey, W. J., & Wickens, C. D. (2004). Driving and side task performance:
The effects of display clutter, separation, and modality. Human Factors, 46(4),
611.

Horrey, W. J., Wickens, C. D., & Consalus, K. P. (2006). Modeling drivers’ vi-
sual attention allocation while interacting with in-vehicle technologies. Journal
of Experimental Psychology: Applied, 12, 67-78.

36
Hosking, S. G., Young, K. L., & Regan, M. A. (2009). The Effects of Text
Messaging on Young Drivers. Human Factors, 51(4), 582-592.

Jamson, A. H., Westerman, S. J., Hockey, G. R. J., & Carsten, O. M. J. (2004).

Speech-based e-mail and driver behaviour: Effects of an in- vehicle message
system interface. Human Factors, 46, 625–639.

Kircher, A., Vogel, K., Tornros, J., Bolling, A., Nilsson, L., Patten, C., Malm-
strom, T., & Ceci, R. (2004). Mobile telephone simulator study. Linkoping,
Sweden: Swedish National Road and Transport Research Institute.

Lansdown, T. C. (2001). Causes, measures, and effects of driver visual work-

load. In P. A. Hancock & P. A. Desmond (Eds.), Stress, workload, and fatigue
(pp. 351–369). Mahwah, NJ: Lawrence Erlbaum.

Lee, J. D., Caven, B., Haake, S., & Brown, T. L. (2001). Speech- based
interaction with in-vehicle computers: The effect of speech-based e-mail on
drivers’ attention to the roadway. Human Factors, 43, 631–640.

Lee, J. D., Regan, M. A., & Young, K. L. (2008). What drives distraction?
Driver distraction as a breakdown of multi-level control. In M. A. Regan, J.
D. Lee, & K. Young (Eds.), Driver distraction: Theory, effects and mitigation
(pp. 41–56). Boca Raton, FL: CRC.

McEvoy, S. P., Stevenson, M. R., & Woodward, M. (2006). The impact of

driver distraction on road safety: Results from a representative survey in two
Australian states. Injury Prevention, 12, 242–247.

One driver in 20 ’uses Facebook at the wheel’. (2010). Retrieved August 25,
2010, from http://www.telegraph.co.uk/motoring/news/7315257/One-driver-
in-20-uses-Facebook-at-the-wheel.html

Salvucci, D.D. (2001). Predicting the effects of in-car interface use on driver
performance: An integrated model approach. International Journal of Human-
Computer Studies, 55, 85-107.

37
Salvucci, D.D., & Macuga, K.L. (2002). Predicting the effects of cellular-phone
dialing on driver performance. Cognitive Systems Research, 3, 95-102.

Seatbelt and mobile phone use surveys: 2009 results. (2010). Retrieved August
25, 2010, from http://www.dft.gov.uk/pgr/statistics/
datatablespublications/accidents/seatbeltmobile

Strayer, D. L., & Johnston, W. A. (2001). Driven to distraction: Dual-task

studies of simulated driving and conversing on a cellular phone. Psychological
Science, 12, 462–466.

Thurlow, C., & Brown, A. (2003). Generation Txt? The sociolinguistics of

young people’s text-messaging. Discourse Analysis Online, 1(1), 30.

Tsimhoni, O., Smith, D., & Green, P. (2004). Address entry while driving:
Speech recognition versus a touch-screen keyboard. Human Factors, 46(4),
600.

38
Appendix A

List of text messages

What do you think of the steering wheel?

What do you think of the touchscreen?

Are you sitting comfortably?

What do you think of your phone? I might get the same one.

What do you think of the course? My brother might apply for next year.

What do you think of the iPad? I might get one next week.

What do you think of your area? I might live there next year.

What do you think of your computer? I’m looking to get a similar one.

What do you think of the new iPhone? I might get one for my birthday.

Can you get me a coffee after this? I’ll be in the common room.

Can you print something for me? I’ve sent it to your email.

Can you get my results for me? I can’t come in to uni.

Can you get me a snack from the shop? I’ll pay you back.

Can you pick up a book from Waterstone’s for me? I ordered it yesterday.

Can you pick up my memory stick? I left it in the 6.08 lab.

What are you doing tomorrow night?

What are you doing later?

39
What are you doing on Sunday?

What are you doing tonight?

What are you doing tomorrow morning?

What are you doing this weekend?

Want to go to a gig next week?

Want to see a movie tonight?

Want to do an experiment together tomorrow?

Want to come to a bbq this weekend?

Want to go for coffee after this?

Want to go for dinner tomorrow night?

How’s your day going?

How’s your week been?

How’s your project going?

How was the journey in today?

How was your weekend?

How was last night?

Where’s the Rockefeller building? I’m late for an experiment!

Where’s the Main Library? I need to print something.

Where’s the Cruciform building? I’m doing an experiment there tomorrow.

Where’s the Psychology building? I want to look at the cubicles.

Where’s the UCL Union? I’ve got a meeting there later.

Where’s the Jeremy Bentham room? I think I’ll work there today.

40