Content uploaded by Anna Frohnwieser
Author content
All content in this area was uploaded by Anna Frohnwieser on Feb 06, 2014
Content may be subject to copyright.
DIPLOMARBEIT
Titel der Diplomarbeit
Human Walking Behavior – The Effect of Density on Walking
Speed and Direction
Verfasserin
Anna Frohnwieser
angestrebter akademischer Grad
Magistra der Naturwissenschaften (Mag.rer.nat.)
Wien, 2012
Studienkennzahl lt. Studienblatt: A 442
Studienrichtung lt. Studienblatt: Biologie / Anthropologie
Betreuerin / Betreuer: Prof. Dr. Karl Grammer
2
Acknowledgements
I dedicate this thesis to my mother, who supported me mentally and financially and
without whom I would not have been able to study and become the person I am today. Danke
Mama!
I want to thank Dr. Elisabeth Oberzaucher for her support and supervision during the
long way from the first idea to this study to the finished thesis. Thanks also to Prof. Karl
Grammer for his constant input!
I am also very grateful to Richard Hopf and the rest of the team at Yellowfish GmbH
for making this study possible and for letting me get an insight into the development of new
software and computer systems. It was great working with you!
Thank you Benno, for always being there for me and for supporting me more than you
even know!
At last I want to thank everyone I didn’t mention who supported me during the last
years – my family, my friends and the numerous people at university who I had the pleasure
to work, chat and drink with.
3
4
Table of content
Acknowledgements .................................................................................................................... 3
Table of content .......................................................................................................................... 5
I. INTRODUCTION .................................................................................................................. 7
1.1. Edward T. Hall – Distances in man ..................................................................... 7
1.2. Personal space ..................................................................................................... 9
1.2.1. Personal space – an evolutionary view ....................................................... 10
1.2.2. The size and shape of personal space ......................................................... 11
1.3. Pedestrian behavior ........................................................................................... 12
1.4. Obstacle circumvention ..................................................................................... 13
1.5. Hypotheses ........................................................................................................ 14
II. METHODS .......................................................................................................................... 15
2.1. CCB Analyser .................................................................................................... 15
2.2. Variables ............................................................................................................ 18
2.3. Data.................................................................................................................... 21
2.3.1. Data transformation. ................................................................................... 22
III. RESULTS ........................................................................................................................... 23
3.1. Descriptive statistics .......................................................................................... 23
3.2. Comparison of weekdays .................................................................................. 23
3.3. Comparison of times .......................................................................................... 25
3.4. Comparison of counter areas ............................................................................. 27
3.5. Density ............................................................................................................... 30
3.6. Direction change ................................................................................................ 31
3.7. Speed ................................................................................................................. 32
3.8. Personal space ................................................................................................... 35
IV. DISCUSSION .................................................................................................................... 39
4.1. Comparison of weekdays .................................................................................. 39
4.2. Comparison of times .......................................................................................... 40
4.3. Comparison of counter areas ............................................................................. 40
4.4. Density ............................................................................................................... 41
4.5. Direction change ................................................................................................ 41
4.6. Speed ................................................................................................................. 42
4.7. Personal space ................................................................................................... 42
4.8. CCB Analyser – future implications ................................................................. 44
V. REFERENCES .................................................................................................................... 45
ZUSAMMENFASSUNG ......................................................................................................... 49
ABSTRACT ............................................................................................................................. 51
Curriculum vitae ....................................................................................................................... 53
5
6
I. INTRODUCTION
Every one of us has made experiences with other people coming too close – in a
crowded subway, a restaurant, or on a busy shopping street. We all know what it feels like to
be touched or shuffled around by strangers, how uncomfortable it is and how much stress it
causes. We feel threatened and disrespected in such situations and try to avoid them as good
as we can. This phenomenon is called crowding and occurs in different contexts, for example
as residential crowding, meaning too many people living in a small space, crowding in work
settings and crowding in transportation (see Karlin et al., 1978). To define what crowding is
we need to know which distances people usually keep from each other in different situations.
In his book “The hidden dimension” (1966) Edward T. Hall defines these distances. What
changes for two people keeping different distances from each other is described by sensory
factors, such as “How much can I see of the other person?” or “Can I smell, touch or hear the
other person?”.
1.1. Edward T. Hall – Distances in man
Intimate distance – close phase. This is the distance where physical contact occurs,
for example when two people hug. At this distance vision is blurred and details of the other
individual can be seen. One can smell, feel and hear the other and sense heat from the other
person’s body.
Intimate distance – far phase (6-8 inches; 15-20cm). One individual can reach the
other with his hands, but the bodies do not touch. Faces are seen as very large and cannot be
perceived as a whole, but every detail is perceptible. Voice is held low and the heat and odor
of the other’s breath can be detected. Crowded subways can bring strangers into intimate
distance; however there are tactics to deal with these unpleasant situations, like being as
7
immobile as possible, trying not to touch another person (and if this is not avoidable keeping
muscles tense), and not looking at another person.
Personal distance – close phase (1.5-2.5 feet; 45-75cm). At this distance one can
hold or grasp the other. Visual distortion of the other’s features is no longer apparent. It
depends on the relationship of the two individuals towards each other if they stay at this
distance or further apart. The better two people know each other and the more they trust each
other the closer they usually get.
Personal distance – far phase (2.5-4 feet; 75-120cm). It extends from a point that is
just outside easy touching distance to a point where two people can touch fingers if they
extend both arms. To perceive the other’s face at this distance, the gaze must wander around
the face since it cannot be seen as a whole. Hands and body of a seated person can be seen.
The voice level is moderate. No body heat is detected, but sometimes the odor of the other
individual can be perceived.
Social distance – close phase (4-7feet; 120-210cm). Greater parts of the face can be
seen at once, and details of skin texture and hair are still perceivable. The head, shoulders, and
upper trunk are visible. Impersonal business occurs at this distance.
Social distance – far phase (7-12 feet; 210-370cm). Business and social discourse at
this distance has a more formal character. Desks in offices keep others at far social distance.
Fine details of the face can not be perceived anymore, but skin texture, hair and clothes are
visible. No heat or odor from the other’s body can be detected. This distance allows
continuing to work or perform other actions when someone is present without being rude.
Public distance – close phase (12-25 feet; 370-760cm). At this distance an individual
can take evasive or defensive action if threatened. No details of the other’s face can be seen,
but the whole body is visible, including peripheral view of other people present.
Public distance – far phase (>25feet; >760cm). 30 feet (915cm) is the distance
automatically set around important public figures, but it can be used by anyone on public
8
occasions. Speech must be loud and slower, and also facial expression is not clearly visible,
so that gestures become more important in communication.
For the study presented in this thesis, the most important distance is personal distance,
also known as personal space.
1.2. Personal space
Personal space is a zone that others are not usually allowed to enter and which
surrounds individuals and in some cases small groups (Cheyne and Efran, 1972). Unwanted
entrances are interpreted as intrusive and cause arousal (Middlemist and Knowles, 1976) and
physical reactions such as increased heart rate, sweating and increased blood pressure.
However, who is allowed to enter and who is not, depends largely on the two individuals.
Persons with whom we have intimate relationships can come closer than strangers, before
discomfort occurs. Personal space is the only human territory that is not bound locally and is
also variable in size depending on the individual and the context. Different personal features
seem to influence which distances create discomfort and which do not. There are differences
in character, for example people who feel confident, powerful and in control keep closer
distances than insecure people (Ickinger and Morris, 2001) and people with Type A
behavioral patterns (characterized by competitive achievement striving, time urgency, and
hostility) who are more responsive to loss of control keep larger distances to others than
people with Type B behavioral patterns (Strube and Werner, 1984). Personal space can also
be influenced by culture. Hall (1966) gives some anecdotes about Europeans keeping closer
distances than North Americans, and he also claims that each cultural group that lives in
North America uses space differently (see Baldassare and Feller, 1975 for a review). Baxter
(1970) showed in his study that Americans of Mexican origin stood very close, Anglo-
Americans kept intermediate distances and African-Americans kept the largest distances from
each other. He also found that female-male groups stayed at the closest distances, female-
9
female groups at intermediate distances and male-male groups kept the biggest interindividual
distances.
People erect boundaries to preserve their personal space – for example pile books
between each other in a library or turn their bodies away from each other in a crowded
subway (Taylor, 1988).
1.2.1. Personal space – an evolutionary view. Why do humans have a natural need to
keep distances from each other and feel discomfort when those distances are disrespected?
Biological explanations are simple. Keeping strangers at a distance that allows
physical contact only with extended arms if at all decreases the possibility of physical
aggression. If you do not know someone or do not trust him it is better to not let him close
enough to touch you, because he might be hostile to you.
Another reason for keeping personal space is to reduce the risk of contagion with
various diseases. When big numbers of people live in a relatively small space and you do not
know most of them anyone could be carrying a contagious disease. Staying away from others
limits the risk of catching a virus by not letting strangers touch you or sneeze or cough in your
direction. For example, the risk of caching influenza increases dramatically for people who
use public transport, where they are forced to stay very close to strangers (Troko et al., 2011).
Personal space also forms around groups of people that belong together in some way,
like friends sitting in a café or co-workers standing next to each other in a hallway. Staying
close to each other and away from others shows that those people belong together and would
probably stick together in any form of confrontation. It is also easier to take care of children
when you are in a group than when you are alone, so safety is another reason for personal
space around groups.
10
People forming groups maintain a certain degree of privacy because others are usually
reluctant to disturb people who are interacting, for example by walking between them
(Cheyne and Efran, 1972).
Personal space can also be described for animals. One example is the observation of
captive polar bears by Renner and Kelly (2006), which showed that these normally solitary
bears always keep a minimum distance from each other, thus limiting aggressive behaviour.
This distance is often referred to as “social distance” or “critical distance”, which is the
minimum distance that reduces aggressive behaviour, for example in cats (Juabe-Díaz, 1999)
or in turkeys reared for commercial use (Buchwalder and Huber-Eicher, 2004).
1.2.2. The size and shape of personal space. Different studies have been conducted
to find out how personal space is shaped and which size it normally has. However, the
findings of these studies vary greatly. For example, Newman and Pollack (1973) found that
personal space is bigger in the rear (~4 feet) than in the front and on the sides (~2 feet each).
They also observed a bigger personal space for deviant adolescents than for normally
behaving ones, indicating a personality factor in the size of personal space.
Leslie Hayduk’s studies contradict Newman and Pollack, since he found that personal
space is bigger in the front than in the rear (Hayduk and Mainprize, 1980; Hayduk, 1981). He
compared the personal space of sighted and blind people and found out that approaching them
creates the same discomfort for both groups at the same distances, indicating equal personal
spaces. In another study (Hayduk, 1981) he investigated personal spaces of sighted test
subjects to find out whether visual mechanisms have an influence on size or shape of personal
space. His study showed that the mean distance for frontal approaches, at which people began
to feel uncomfortable, was 59cm, which would be within the close phase of personal space
(Hall, 1966). He found that personal space is noncircular, with a decline from a maximum
frontal distance to a minimum rear distance. When people were not allowed to move their
11
heads the use of peripheral vision increased the size of personal space at these positions,
making the largest distance occur half-way between the head and body front. Also, rear
distances dropped to almost zero, since people were not allowed to look behind them and
could not see the experimenter.
As described above there are many different study results about the size and shape of
personal space, so that it seems to be very variable. In their 1984 study, Strube and Werner
observed different sizes in personal space for people with different behaviour patterns (Type
A – competitive, hostile, in need for control over the environment and Type B – defined by
the absence of Type A characteristics) and for people in different situations. Their results
show that Type A people in a customer situation preserve the biggest personal space whereas
Type B people in a salesperson situation have the smallest personal space. This is just one of
the many settings that make the concept of personal space very important.
The need for privacy and distance to others also influences which seats we take in a
restaurant (Robson, 2008) and on trains (Evans and Wener, 2007) and how groups of people
sit down at a table (Thompson et al., 1979).
Contrary to these studies, which all describe personal space for standing or sitting
people, the present thesis deals with the personal space of pedestrians. This is why we first
have to take a look at the typical behavior of pedestrians:
1.3. Pedestrian behavior
There is a small set of certain rules that every walking person follows unconsciously.
Pedestrians always like to take the fastest route to their destination (Fajen and Warren, 2003),
even if that route is crowded (Helbing et al., 2001) and take small detours if that allows them
to walk faster (Ganem, 1998).
12
Each person has a desired natural walking speed, dependent among other factors on
age, constitution and sex, which is the least energy-consuming speed for that person (Ralston,
1958). The mean comfortable walking speed is different for different groups of people (e.g.
old and young people, women and men, etc.), and lies usually between 1.2m/s and 1.6m/s
(e.g. Bohannon, 1997; Costa, 2010; Henderson and Lyons, 1972; Vallis and McFadyen,
2003). However, if a person needs to get to a destination within a short amount of time
walking speed increases (Helbing et al., 2001).
Pedestrians like to keep a certain distance between each other (Costa, 2010;
Johansson, 2009) and to objects (Moussaïd et al., 2011) and change their walking speed to
maintain those distances (Konečni et al., 1975), They are also very good in finding the fastest
path through a field of obstacles (Gérin-Lajoie et al., 2007).
Since other people can be considered as obstacles when we are walking, it is important
to know how pedestrians avoid obstacles.
1.4. Obstacle circumvention
Gérin-Lajoie and his colleagues have done studies on obstacle circumvention,
investigating which distances people keep from objects and other people. The shape of these
distances combined looked a bit like a house – with a maximum distance in the front, smaller
distances to each side and a rear distance of zero. In one of their studies (Gérin-Lajoie et al.,
2006) they compared older and younger adults walking past stationary or moving obstacles.
The test subjects walked slower when the obstacle moved and when they did not know the
movement’s direction beforehand. Older adults preserved a larger personal space to the
objects as younger adults. In a similar study (Gérin-Lajoie et al., 2008) the researches found
out that personal space kept between the test subjects and the obstacle did not change with
walking speed, and that personal space was bigger on the left side of the test subjects than on
13
the right. A reason for this may be that most test subjects were right-handed and kept a
smaller safety margin on their dominant side.
1.5. Hypotheses
This present study aims to combine the investigation of pedestrian behaviour with the
investigation of personal space. As shown in studies about the avoidance of obstacles (e.g.
Gérin-Lajoie et al., 2006; Gérin-Lajoie et al., 2008), moving people always keep minimum
distances to objects and to others. However, it has never been shown which measures
pedestrians take to avoid situations in which these minimum distances might be invaded. We
used a system named CCB Analyser to observe general pedestrian behaviour and to pre-
calculate possible invasions into personal space, thus making it possible to examine the
effects of different densities and interpersonal distances on pedestrians.
According to the study design the hypotheses were:
1 Higher density of people and lower interpersonal distances will change walking speed,
most probably in a U-shaped way: at low density/high distances they will walk at
normal speed, at moderate density/distances speed will increase because of stress
(Konečni et al., 1975) and at high density/low distances speed will decrease again due
to the lack of space.
2 Higher density of people and lower interpersonal distances will increase the need to
change directions while walking.
14
II. METHODS
2.1. CCB Analyser
The study was conducted using a system called CCB Analyser, which was developed
by the Viennese company Yellowfish GmbH (Fig.1). The system consisted of five sensors
placed on the ceilings of the Gasometer shopping center, which took photos of the floor every
0.5 seconds. These photos were then compared by the software to find out if a person had
changed his/her position from one picture to the next. This way, the system can track people
through an observed area, making it possible to measure speed, density and other parameters.
The software codes all of the measured parameters into colors in different scales, so that
differences and changes can be easily detected. Also, there is the option to see all the
measured values in numbers in a 50cm x 50cm grid throughout the observed area (Fig. 2).
The program gives out all objects that did not move within the observation time as a
background picture, usually showing the floor and furniture in the area for easier orientation.
It is also possible to define so called counter areas, which are rectangular spaces within the
observed area. Different parameters are calculated by the program for each counter area; the
most important being number of people passing through.
For this study, square counter areas of one square meter in size were used, each
including four grid squares (Fig.2). Three counter areas were placed at different areas of the
Gasometer outline. The places were chosen for being different in terms of density and of
directions pedestrians usually walk there. Counter area 1 was located in a corridor
approximately 5 meter wide with a restaurant on one side (Fig.1), which was used frequently
by people walking either from the left to the right or from the right to the left. Counter area 2
was located at the end of another corridor, where it gets narrower before it opens into a hall.
People there walked in different directions and took different turns to get to their destinations.
Counter area 3 was placed at a crossing, where people could come from three different
directions.
15
Fig. 1: CCB Analyser. The yellow lines indicate the outline of the Gasometer shopping centre;
the grey areas are the background pictures showing floors and furniture; arrows point at the
three counter areas.
CA1 CA2
CA3
16
The observation time can be chosen within the program, from a minimum of five
minutes to a maximum of one day. All parameters are calculated for the desired time span and
can thus be compared for different times. In this study the observed time span was always 30
minutes. For each day included in the data, three times were used – 08:30 to 09:00, 13:00 to
13:30 and 16:00 to 16:30 CET respectively CEST.
Fig. 2: Counter area 1 with values for frequency. Only the four values within the counter area
were recorded.
17
2.2. Variables
Not all measurements of the system were used for this study. The variables which
were used are:
- Total people passing: Number of all people passing through the counter area within
the observed time span.
- Frequency: Number of people per square meter during the observed time span.
- Average speed: Average speed of all people in m/s for the observed time span.
- Average change of direction: Average angle of direction change, including everyone
who changed their direction for more than 10 and less than 90 degrees.
- Number of direction changes: Number of direction changes for more than 10 and less
than 90 degrees during the observed time span.
- Average speed decrease: Average decrease of speed in m/s, including everyone who
decreased their speed for more than 0.25m/s.
- Number of speed decreases: Number of speed decreases for more than 0.25 m/s during
the observed time span.
- Average speed increase: Average increase of speed in m/s, including everyone who
increased their speed for more than 0.25m/s.
- Number of speed increases: Number of speed increases for more than 0.25 m/s during
the observed time span.
18
Since the main purpose of this study was to investigate the personal space of
pedestrians, we had to find ways to measure violations of pedestrians’ personal spaces. To do
this two different approaches were used. The first one was named personal space in circles
(from now on referred to as PSC): Around every person in one picture recorded by the sensors
circles of different diameters were drawn. These circles represented different personal space
areas around the person. When there was another person within these circles, points were
added up depending on how close (e.g. within which circle) this person was. All the points for
each person in each picture over the observed time span were added for each grid square. An
example for this calculation is shown in Fig. 3.
Circle diameters and respective points:
5-10m = 1 point
1-5m = 2 points
<1m = 3 points
Fig. 3: Personal space in circles (PSC). The black dot marks a person, the circles around it the
three different personal space distances. The grey dots mark other people located near the
observed person. The calculation for the PSC value of this person would be: 2 + 2 + 1 = 5
19
The second approach included movements of the observed people. Based on Gérin-
Lajoie et al.’s study (2007), a “house-shaped” personal space was chosen, being bigger in the
front and zero in the rear. The distance to each side was 80cm and to the front 160cm. For
each person in one picture, the program calculated the positions of this person in the next
pictures, assuming that the person continued walking in the exact same direction as he/she did
until then. If another person entered the first person’s personal space within one of the next
pictures, one point was added to the first person (see Fig. 4 for an example). This was done
for every person in each picture, and all the points for each square grid and within the
observed time span were added.
The size of the personal space could be enlarged without changing the shape or
proportions. Also, the number of frames being calculated ahead could be changed.
Therefore, we calculated six different personal space values for this study:
1. Planning 2 frames (≙1 second) ahead with a personal space of 80cm to the sides and
160cm to the front. From now on referred to as PS2/small.
2. Planning 4 frames (≙2 seconds) ahead with a personal space of 80cm to the sides and
160cm to the front. From now on referred to as PS4/small.
3. Planning 6 frames (≙3 seconds) ahead with a personal space of 80cm to the sides and
160cm to the front. From now on referred to as PS6/small.
4. Planning 2 frames (≙1 second) ahead with a personal space of 120cm to the sides and
240cm to the front. From now on referred to as PS2/big.
5. Planning 4 frames (≙2 seconds) ahead with a personal space of 120cm to the sides and
240cm to the front. From now on referred to as PS4/big.
20
6. Planning 6 frames (≙3 seconds) ahead with a personal space of 120cm to the sides and
240cm to the front. From now on referred to as PS6/big.
2.3. Data
The data used for this study were recorded from the 22nd of June 2010 to the 13th of
November 2010. There was no data used between the 15th of July 2010 and the 11th of
October 2010, because one of the sensors did not work during that period.
Fig. 4: Personal space in motion. The black dot marks the observed person. In the left picture
the lines show the shape of the personal space, being bigger in the front and zero in the rear. The
picture on the right shows a person at a given moment (bigger dot) with two positions/frames of
the same person being calculated ahead (smaller dots). The gray arrow with grey dots shows a
second person’s path. At frame 2, the grey path crosses the observed person’s personal space.
Therefore one point is counted for the observed person.
21
Altogether, 507 datasets were recorded, including 19 variables each. Every set
contains a 30 minute time span. For each day, 9 sets were recorded – at three different times
and three different counter areas.
2.3.1. Data transformation. Since there were four values for each variable (excluding
the total number of people passing the counter area) because of the four grid squares within
the counter area and because many variables were dependent on the number of people, those
variables had to be transformed: For frequency, the sum of the four values was calculated. For
average speed, average change of direction, average speed decrease and average speed
increase the average of the four values was calculated. For number of direction changes,
number of speed decreases and number of speed increases the percentages of people changing
direction or speed were calculated. For all personal space values the sums of the four values
were calculated and then divided by the total number of people crossing the area.
22
III. RESULTS
3.1. Descriptive statistics
Altogether 15 210 minutes (= 10.56 days) of recordings were used for this study,
divided into 30 minute sets. 66 sets were recorded on Mondays (13.0%), 81 on Tuesdays and
Wednesdays (16.0% each), 72 on Thursdays, Fridays and Saturdays (14.2% each) and 63 on
Sundays (12.4%). 168 sets were recorded in the morning and at midday (33.1% each), while
171 were recorded on afternoons (33.7%). Each counter area was recorded 169 times (33.3%).
On average 49.83 (range: 0-173; SD: 35.08) people were observed in each set. This
makes altogether 25 265 observed people. The mean frequency (people per square meter) was
169.75 (range: 0-604; SD: 123.07). The average speed of the observed people was 1.71m/s
(range: 0-3.28m/s; SD: 0.52). On average 15.1% (range: 0-100%; SD: 9.71) of the observed
people decreased their speed, with an average speed decrease of 0.31m/s (range: 0-0.83m/s;
SD: 0.17). 14.5% of the observed people increased their speed (range: 0-66.67%; SD: 8.81)
with an average speed increase of 0.31m/s (range: 0-0.88m/s; SD: 0.16). The average
direction change was 15.85 degrees (range: 0-43.25; SD: 7.98), with 29.66% of people
changing their direction for more than 10 degrees.
The mean value for the circular personal space (PSC) was 1.95 (range: 0-8.34; SD:
1.51), for PS2/small 0.04 (range: 0-0.35; SD: 0.05), PS4/small (range: 0-0.63; SD: 0.09), for
PS6/small 0.11 (range: 0-0.79; SD: 0.12), PS2/big 0.09 (range: 0-0.58; SD: 0.11), PS4/big
0.15 (range: 0-0.75; SD: 0.16), and for PS6/big 0.2 (range: 0-0.94; SD: 0.21).
3.2. Comparison of weekdays
When comparing weekdays almost no differences were found between workdays
(Monday to Friday). The average direction change was higher on Tuesdays than on Mondays
(Z=-2.071; p=0.038) and the average speed was higher on Tuesdays than on Thursdays (Z=-
23
2.159; p=0.031). All other differences were found between workdays and weekends (Saturday
and Sunday); therefore further analysis was conducted comparing all workdays with
weekends (Table 1). The total number of observed people and the frequency of people were
higher on workdays (Fig.5), as were average speed, average speed decrease, average speed
increase and average direction change. There were less invasions of circular personal space
per person on weekends than during the week (Fig. 6).
There were no differences in the percentages of people changing direction or speed,
and also in the six personal space values PS2/small to PS6/big.
Table 1: Comparison of weekdays (Monday to Friday) versus weekends (Saturday and
Sunday).
Total = average total number of observed people per 30min. Fr = frequency of observed
people per square meter. Speed = average walking speed. Angle = average angle of
direction change. Angle% = percentage of people changing direction. SpeedDE = average
speed decrease in m/s. SpeedDE% = percentage of people decreasing speed. SpeedIN =
average speed increase in m/s. SpeedIN% = percentage of people increasing speed. PSC =
circular personal space. PS2/small – PS6/big = personal space in motion (see above for
detailed explanation).
Weekdays (N=370) Weekends (N=131) Mann-Whitney U test
Median Median Z-Value p-Value
Total 51.00 23.00 -8.529 <0.001
Fr 172.00 76.00 -8.366 <0.001
Speed 1.55 1.47 -4.474 <0.001
Angle 17.75 14.00 -5.163 <0.001
Angle% 26.97 29.79 n.s
SpeedDE 0.33 0.23 -5.777 <0.001
SpeedDE% 13.11 14.81 n.s
SpeedIN 0.33 0.23 -5.794 <0.001
SpeedIN% 13.22 13.33 n.s
PSC 1.56 1.25 -3.387 0.001
PS2/small 0.03 0.00 n.s
PS4/small 0.05 0.03 n.s
PS6/small 0.06 0.05 n.s
PS2/big 0.06 0.03 n.s
PS4/big 0.09 0.08 n.s
PS6/big 0.13 0.13 n.s
24
sunsatfrithurswedtuesmon
Weekday
80
60
40
20
0
95% CI Total Number of observed People
Number of observed People compared by Weekday
3.3. Comparison of times
The three observation times – 08:30 to 09:00, 13:00 to 13:30 and 16:00 to 16:30 –
were compared (see Table 2). Analysis shows that there were great differences, especially
when comparing morning to midday and afternoon. The total number of observed people and
the frequency of people were lower in the morning than later during the day (Fig. 7). Average
direction change, average speed decrease and average speed increase were also lower, as were
the percentages of people decreasing their speed and changing direction. Personal space was
less invaded in the morning (Fig. 8). There were no differences in the average speed and the
percentage of people increasing their speed between morning, midday and afternoon.
There were more people observed during midday than in the afternoon and the average
direction change and speed increase were higher. All other variables did not differ between
midday and afternoon.
Fig. 6: Circular personal space compared by weekday.
More violations of circular personal space occurred
during the workweek than on weekends.
Fig. 5: Number of observed people compared by
weekday. More people were observed during the
workweek than on weekends.
sunsatfrithurswedtuesmon
Weekday
2,50
2,00
1,50
1,00
95% CI Circular Personal Space
Circular Personal Space compared by Weekday
25
Table 2: Comparison of different times of day. Total = average total number of observed people per 30min. Fr = frequency of observed people per square meter.
Speed = average walking speed. Angle = average angle of direction change. Angle% = percentage of people changing direction. SpeedDE = average speed
decrease in m/s. SpeedDE% = percentage of people decreasing speed. SpeedIN = average speed increase in m/s. SpeedIN% = percentage of people increasing
speed. PSC = circular personal space. PS2/small – PS6/big = personal space in motion (see above for detailed explanation).
Mann-Whitney U test
08:30 (N=162) 13:00 (N=168) 16:00 (N=171) 08:30 / 13:00 08:30 / 16:00 13:00 / 16:00
Median Median Median Z-Value p-Value Z-Value p-Value Z-Value p-Value
Total 20.50 64.00 48.00 -10.987 <0.001 -9.512 <0.001 -4.036 <0.001
Fr 68.00 213.00 156.00 -10.768 <0.001 -9.134 <0.001 -3.948 <0.001
Speed 1.55 1.54 1.51 n.s. n.s. n.s.
Angle 11.63 19.13 16.75 -8.345 <0.001 -5.939 <0.001 -3.832 <0.001
Angle% 23.76 32.35 28.97 -3.952 <0.001 -2.399 0.016 n.s.
SpeedDE 0.18 0.35 0.35 -8.032 <0.001 -7.719 <0.001 n.s.
SpeedDE% 11.76 14.04 14.55 -2.135 0.033 -2.080 0.038 n.s.
SpeedIN 0.20 0.35 0.33 -7.786 <0.001 -6.380 <0.001 -2.345 0.019
SpeedIN% 12.50 13.65 13.21 n.s. n.s. n.s.
PSC 0.82 1.80 1.87 -8.961 <0.001 -8.153 <0.001 n.s.
PS2/small 0.00 0.03 0.03 -5.445 <0.001 -5.070 <0.001 n.s.
PS4/small 0.00 0.05 0.06 -5.626 <0.001 -5.065 <0.001 n.s.
PS6/small 0.03 0.08 0.08 -5.943 <0.001 -5.466 <0.001 n.s.
PS2/big 0.00 0.07 0.07 -6.482 <0.001 -5.867 <0.001 n.s.
PS4/big 0.03 0.12 0.10 -6.324 <0.001 -5.700 <0.001 n.s.
PS6/big 0.06 0.16 0.14 -6.600 <0.001 -5.896 <0.001 n.s.
26
16:00 - 16:3013:00 - 13:3008:30 - 09:00
Time
80
70
60
50
40
30
20
95% CI Number of observed People
Number of observed People compared by Time of Day
16:00 - 16:3013:00 - 13:3008:30 - 09:00
Time
0,30
0,25
0,20
0,15
0,10
95% CI PS6/big
Personal Space Violations compared by Time of Day
3.4. Comparison of counter areas
Since the three counter areas were chosen for being different in terms of possible walking
directions, all the variables were compared between them (see Table 3).
Every variable differed between counter areas 1 and 2. The total number of observed
people and the frequency were higher in counter area 1 (Fig. 9), while average speed (Fig. 10),
average speed decrease and average speed increase were lower, as were the percentages of
people changing their speed. Average direction change and percentage of people changing
direction were higher in counter area 1 than in counter area 2, and there were also more
violations of personal space (Fig. 11).
The total number of observed people (Fig. 9), the frequency and the average speed (Fig.
10) were higher in counter area 1 than in counter area 3.The average direction change and the
percentage of people changing direction were lower. The average speed decrease was the same
between the two counter areas, but the percentage of speed decreases was higher in counter area
Fig. 8: Personal space (PS6/big) compared by time
of day. Least violations of personal space occurred
in the morning, most during midday.
Fig. 7: Number of observed people compared by
time of day. Least people were observed in the
morning, most during midday.
27
3. Average speed increase and percentage of speed increases were both higher in counter area 1.
There were more violations of personal space in counter area 1 than in counter area 3 (Fig. 11).
When comparing counter areas 2 and 3 the following differences were found: the total
number of observed people and the frequency were higher in counter area 2 (Fig. 9), as were
average speed (Fig. 10) and average speed decrease and increase. Also the percentage of people
increasing their speed was higher. Average direction change and percentage of people changing
direction were higher in counter area 3. Percentage of people decreasing their speed was the
same for the two counter areas. There were more violations of circular personal space and of
PS2/small in counter area 3, whereas the other personals space measures did not differ (Fig. 11).
28
Table 3: Comparison of the three counter areas. Total = average total number of observed people per 30min. Fr = frequency of observed people per square meter.
Speed = average walking speed. Angle = average angle of direction change. Angle% = percentage of people changing direction. SpeedDE = average speed
decrease in m/s. SpeedDE% = percentage of people decreasing speed. SpeedIN = average speed increase in m/s. SpeedIN% = percentage of people increasing
speed. PSC = circular personal space. PS2/small – PS6/big = personal space in motion (see above for detailed explanation).
Mann-Whitney U test
CA1 (N=168) CA2 (N=169) CA3 (N=164) CA1 / CA2 CA1 / CA3 CA2 / CA3
Median Median Median Z-Value p-Value Z-Value p-Value Z-Value p-Value
Total 76.00 43.00 30.00 -7.247 <0.001 -9.710 <0.001 -4.360 <0.001
Fr 266.00 140.00 96.00 -8.286 <0.001 -10.270 <0.001 -4.023 <0.001
Speed 1.52 2.30 1.42 -14.008 <0.001 -7.186 <0.001 -14.531 <0.001
Angle 16.25 13.25 19.00 -2.520 0.012 -4.132 <0.001 -4.149 <0.001
Angle% 25.99 12.24 47.49 -9.523 <0.001 -12.191 <0.001 -14.051 <0.001
SpeedDE 0.30 0.45 0.28 -8.076 <0.001 n.s. -8.291 <0.001
SpeedDE% 10.23 17.24 15.87 -8.853 <0.001 -7.976 <0.001 n.s.
SpeedIN 0.30 0.43 0.23 -7.819 <0.001 -3.012 0.003 -8.949 <0.001
SpeedIN% 10.41 16.67 13.33 -8.833 <0.001 -3.445 0.001 -4.509 <0.001
PSC 3.33 1.00 1.48 -12.472 <0.001 -9.446 <0.001 -6.087 <0.001
PS2/small 0.07 0.00 0.00 -11.106 <0.001 -9.195 <0.001 -2.131 0.033
PS4/small 0.17 0.02 0.02 -12.406 <0.001 -11.861 <0.001 n.s.
PS6/small 0.24 0.04 0.03 -12.806 <0.001 -12.703 <0.001 n.s.
PS2/big 0.18 0.03 0.03 -12.121 <0.001 -10.858 <0.001 n.s.
PS4/big 0.32 0.05 0.04 -13.059 <0.001 -12.653 <0.001 n.s.
PS6/big 0.43 0.08 0.07 -13.089 <0.001 -12.876 <0.001 n.s.
29
321
Counter Area
100
80
60
40
20
95% CI Number of observed People
Number of observed People compared by Counter Area
321
Counter Area
2,25
2,00
1,75
1,50
1,25
95% CI Average Speed m/s
Average Speed compared by Counter Area
321
Counter Area
0,40
0,30
0,20
0,10
95% CI PS6/big
Personal Space Violations compared by Counter Area
3.5. Density
Density, measured by the number of observed people and the frequency of people,
correlates with almost every other variable (Table 7). The more people there were at the same
Fig. 9: Number of observed people compared by
counter area. Most people were observed in counter
area 1, least in counter area 3.
Fig. 10: Average speed compared by counter
area. People were fastest in counter area 2.
Fig. 11: Personal space (PS6/big) compared by counter area. Most violations of personal space
occurred in counter area 1. There was no significant difference between counter areas 2 and 3.
30
time the faster they walked (Fig. 12) and the more they changed direction and speed. There
were also more violations of personal space at higher densities (Fig. 13).
200150100500
Number of observed People per 30min
4,00
2,00
0,00
Average Speed m/s
Correlation of Number of observed People and Average Speed
200150100500
Number of observed People per 30min
1,00
0,80
0,60
0,40
0,20
0,00
PS6/big
Correlation of Number of observed People and Personal Space Violations
3.6. Direction change
The average degree of change of direction correlates with every other variable except
the percentage of people increasing their speed (Table 7). The average degree to which
pedestrians changed their direction was higher with higher density (Fig. 14) and lower
average walking speed. When higher direction changes occurred there were also more
changes of walking speed and more violations of personal space.
The percentage of people changing their direction does not correlate with density. It
was higher at lower walking speeds and when changes of speed were smaller. More people
changed direction when the number of violations of circular personal space was higher (Fig.
15), but there was no correlation with personal space in motion (PS2/small – PS6/big).
Fig. 12: Correlation of number of observed people and
average walking speed. rs = 0.494
Fig. 13: Correlation of number of observed people and
personal space violations (PS6/big). rs = 0.659
31
200150100500
Number of observed People per 30min
40,00
20,00
0,00
Average Angle of Direction Change
Correlation of Number of observed People and Direction Change
140,00120,00100,0080,0060,0040,0020,000,00
Percentage of People changing Direction
10,00
8,00
6,00
4,00
2,00
0,00
Circular Personal Space
Correlation of Percentage of People changing Direction and Violations of Circular
Personal Space
3.7. Speed
People walked faster at higher densities. There were smaller and less direction changes
and bigger changes of speed at higher average walking speed. At higher speed there were less
violations of circular personal space (Fig. 16). No correlations could be found with personal
space in motion (PS2/small – PS6/big) (Table 7).
The average speed decrease and the average speed increase correlate positively with
almost every other variable. The only exception is the percentage of people changing
direction, which becomes smaller as changes of speed become bigger.
The percentage of people decreasing their speed correlates with every variable except
for average speed. More people decreased their speed at lower densities and when there were
less violations of personal space.
The percentage of people increasing their speed correlates with every variable except
for the number of observed people, frequency, average change of direction, and percentage of
people changing direction. It shows the same pattern as the percentage of people decreasing
their speed, being bigger when less violations of personal space occur. The changes of
Fig. 14: Correlation of number of observed people and
average angle of direction change. rs = 0.428
Fig. 15: Correlation of percentage of people changing
direction and violations of circular personal space.
rs = 0.230
32
direction and speed including the degree of changes and the percentage of people changing
their walking behavior all correlate positively.
3,503,002,502,001,501,000,500,00
Average Speed m/s
10,00
8,00
6,00
4,00
2,00
0,00
Circular Personal Space
Correlation of Average walking Speed and Violations of Circular Personal Space
Since average speed was higher in counter area 2 than in the other counter areas (Fig.
10), all correlations were calculated excluding counter area 2 (Table 5). In this analysis
average speed correlated with every variable – including number of observed people (Fig. 17)
and all the personal space values (Fig. 18) – except for the average direction change and the
percentage of people increasing their speed.
Fig. 16: Correlation of average walking speed and
violations of circular personal space. rs = -0.150
33
Table 5: Correlation of average walking speed, excluding counter area 2. Total = average total number of
observed people per 30min. Fr = frequency of observed people per square meter. Speed = average walking
speed. Angle = average angle of direction change. Angle% = percentage of people changing direction. SpeedDE
= average speed decrease in m/s. SpeedDE% = percentage of people decreasing speed. SpeedIN = average
speed increase in m/s. SpeedIN% = percentage of people increasing speed. PSC = circular personal space.
PS2/small – PS6/big = personal space in motion (see above for detailed explanation).
Total Fr Speed Angle
Spearman's rho Speed Correlation Coefficient 0.494 0.502 1.000 -0.012
Sig. (2-tailed) <0.001 <0.001 . 0.822
N 338 338 338 338
Angle% SpeedDE SpeedDE% SpeedIN
Speed Correlation Coefficient -0.480 0.293 -0.190 0.282
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001
N 332 338 332 338
SpeedIN% PSC PS2/small PS4/small
Speed Correlation Coefficient -0.092 0.388 0.386 0.378
Sig. (2-tailed) 0.095 <0.001 <0.001 <0.001
N 332 332 332 332
PS6/small PS2/big PS4/big PS6/big
Speed Correlation Coefficient 0.369 0.403 0.406 0.389
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001
N 332 332 332 332
200150100500
Number of observed People
2,50
2,00
1,50
1,00
0,50
0,00
Average Speed m/s
Correlation of Number of observed People and average walking Speed without
Counter Area 2
1,00
0,80
0,60
0,40
0,20
0,00
PS6/big
2,001,501,000,50
Average Speed m/s
Correlation of average walking Speed and Violations of Personal Space
Fig. 17: Correlation of number of observed people and
average walking speed. rs = 0.494. Counter Area 2
was excluded, because of the unusually high average
walking speed that was measured there.
Fig. 18: Correlation of average walking speed and
violations of personal space (PS6/big). rs = 0.389.
Counter Area 2 was excluded, because of the
unusually high average walking speed that was
measured there.
34
Speed and the personal space values were also correlated partially, controlling for the
total number of people observed. Here, average speed correlated negatively with every
personal space value (Table 6).
Table 6: Partial correlation of average walking speed with personal space, controlling for number of people. Total
= average total number of observed people per 30min. Speed = average walking speed. PSC = circular personal
space. PS2/small – PS6/big = personal space in motion (see above for detailed explanation).
Control Variables PSC PS2/small PS4/small PS6/small
Total Speed Correlation Coefficient -0.430 -0.189 -0.252 -0.261
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001
df 498 498 498 498
PS2/big PS4/big PS6/big
Speed Correlation Coefficient -0.236 -0.269 -0.267
Sig. (2-tailed) <0.001 <0.001 <0.001
df 498 498 498
3.8. Personal space
Circular personal space (PSC) correlates with every other variable (Table 7). More
violations occurred at higher density and lower speed. The more violations of circular
personal space occurred the more people changed walking direction, and there were less but
bigger changes of speed. The six other personal space values (PS2/small – PS6/big) show the
same pattern, but do not correlate with average speed and percentage of people changing
direction.
35
Table 7: Correlations of all variables. Total = average total number of observed people per 30min. Fr = frequency of observed people per square meter. Speed = average walking
speed. Angle = average angle of direction change. Angle% = percentage of people changing direction. SpeedDE = average speed decrease in m/s. SpeedDE% = percentage of people
decreasing speed. SpeedIN = average speed increase in m/s. SpeedIN% = percentage of people increasing speed. PSC = circular personal space. PS2/small – PS6/big = personal space
in motion (see above for detailed explanation).
Total Fr Speed Angle Angle% SpeedDE SpeedDE% SpeedIN SpeedIN%
Spearman's rho Total Correlation Coefficient 1.000 0.994 0.231 0.428 -0.045 0.530 -0.113 0.548 -0.073
Sig. (2-tailed) . <0.001 <0.001 <0.001 0.314 <0.001 0.012 <0.001 0.101
N 507 507 507 507 501 507 501 507 501
Fr Correlation Coefficient 0.994 1.000 0.207 0.423 -0.032 0.507 -0.122 0.528 -0.081
Sig. (2-tailed) <0.001 . <0.001 <0.001 0.475 <0.001 0.006 <0.001 0.069
N 507 507 507 507 501 507 501 507 501
Speed Correlation Coefficient 0.231 0.207 1.000 -0.144 -0.690 0.465 0.037 0.482 0.211
Sig. (2-tailed) <0.001 <0.001 . 0.001 <0.001 <0.001 0.411 <0.001 <0.001
N 507 507 507 507 501 507 501 507 501
Angle Correlation Coefficient 0.428 0.423 -0.144 1.000 0.484 0.380 0.218 0.323 0.087
Sig. (2-tailed) <0.001 <0.001 0.001 . <0.001 <0.001 <0.001 <0.001 0.053
N 507 507 507 507 501 507 501 507 501
Angle% Correlation Coefficient -0.045 -0.032 -0.690 0.484 1.000 -0.133 0.186 -0.189 -0.045
Sig. (2-tailed) 0.314 0.475 <0.001 <0.001 . 0.003 <0.001 <0.001 0.311
N 501 501 501 501 501 501 501 501 501
SpeedDE Correlation Coefficient 0.530 0.507 0.465 0.380 -0.133 1.000 0.443 0.636 0.189
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 0.003 . <0.001 <0.001 <0.001
N 507 507 507 507 501 507 501 507 501
SpeedDE% Correlation Coefficient -0.113 -0.122 0.037 0.218 0.186 0.443 1.000 0.107 0.131
Sig. (2-tailed) 0.012 0.006 0.411 <0.001 <0.001 <0.001 . 0.017 0.003
N 501 501 501 501 501 501 501 501 501
SpeedIN Correlation Coefficient 0.548 0.528 0.482 0.323 -0.189 0.636 0.107 1.000 0.484
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.017 . <0.001
N 507 507 507 507 501 507 501 507 501
SpeedIN% Correlation Coefficient -0.073 -0.081 0.211 0.087 -0.045 0.189 0.131 0.484 1.000
Sig. (2-tailed) 0.101 0.069 <0.001 0.053 0.311 <0.001 0.003 <0.001 .
N 501 501 501 501 501 501 501 501 501
36
PSC PS2/small PS4/small PS6/small PS2/big PS4/big PS6/big
Spearman's rho Total Correlation Coefficient 0.749 0.559 0.591 0.608 0.631 0.653 0.659
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
Fr Correlation Coefficient 0.769 0.580 0.615 0.634 0.655 0.680 0.686
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
Speed Correlation Coefficient -0.150 -0.022 -0.030 -0.007 -0.014 -0.004 0.018
Sig. (2-tailed) 0.001 0.620 0.505 0.873 0.760 0.936 0.694
N 501 501 501 501 501 501 501
Angle Correlation Coefficient 0.395 0.224 0.190 0.170 0.231 0.202 0.195
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
Angle% Correlation Coefficient 0.230 0.066 0.008 -0.023 0.057 0.009 -0.014
Sig. (2-tailed) <0.001 0.139 0.859 0.606 0.205 0.834 0.754
N 501 501 501 501 501 501 501
SpeedDE Correlation Coefficient 0.223 0.161 0.137 0.137 0.182 0.177 0.180
Sig. (2-tailed) <0.001 <0.001 0.002 0.002 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
SpeedDE% Correlation Coefficient -0.133 -0.169 -0.217 -0.220 -0.190 -0.218 -0.213
Sig. (2-tailed) 0.003 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
SpeedIN Correlation Coefficient 0.206 0.117 0.102 0.136 0.157 0.154 0.172
Sig. (2-tailed) <0.001 0.009 0.023 0.002 <0.001 0.001 <0.001
N 501 501 501 501 501 501 501
SpeedIN% Correlation Coefficient -0.183 -0.171 -0.184 -0.186 -0.152 -0.174 -0.183
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
37
PSC PS2/small PS4/small PS6/small PS2/big PS4/big PS6/big
Spearman's rho PSC Correlation Coefficient 1.000 0.715 0.717 0.725 0.788 0.780 0.777
Sig. (2-tailed) . <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
PS2/small Correlation Coefficient 0.715 1.000 0.875 0.812 0.879 0.822 0.776
Sig. (2-tailed) <0.001 . <0.001 <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
PS4/small Correlation Coefficient 0.717 0.875 1.000 0.939 0.858 0.927 0.883
Sig. (2-tailed) <0.001 <0.001 . <0.001 <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
PS6/small Correlation Coefficient 0.725 0.812 0.939 1.000 0.840 0.919 0.940
Sig. (2-tailed) <0.001 <0.001 <0.001 . <0.001 <0.001 <0.001
N 501 501 501 501 501 501 501
PS2/big Correlation Coefficient 0.788 0.879 0.858 0.840 1.000 0.930 0.888
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 . <0.001 <0.001
N 501 501 501 501 501 501 501
PS4/big Correlation Coefficient 0.780 0.822 0.927 0.919 0.930 1.000 0.962
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 . <0.001
N 501 501 501 501 501 501 501
PS6/big Correlation Coefficient 0.777 0.776 0.883 0.940 0.888 0.962 1.000
Sig. (2-tailed) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 .
N 501 501 501 501 501 501 501
38
IV. DISCUSSION
The goal of this study was to find out more about human walking behavior, especially
about the effect of high density and intrusion into personal space on speed and direction
change. Pedestrians were observed and a computer program called CCB Analyser calculated
averages for speed, density, change of speed, change of direction and two different
approaches to personal space invasions, one being the personal space represented by circles
around a person in single frames and the other one being calculated ahead dependent on the
paths people were walking.
Results show that almost all variables depend on each other. A reason for that could be
one main factor influencing most of the variables – most probably the total number of people.
Even when accounting for it by dividing the variables by the number of observed people or
calculating percentages, these influences persist. The more people there are the higher the
density and therefore the more intrusions into personal space can occur.
4.1. Comparison of weekdays
When comparing weekdays most differences were found between workdays and
weekends. The stores of the Gasometer are closed on Sundays as are the many office
buildings surrounding it. There might be completely different people walking through the
building on weekends, who for example visit the cinema located in Gasometer E or the fitness
studio in Gasometer D. Also, fewer people were observed on weekends than during the
workweek. They walked slower and there were less invasions of circular personal space.
What is remarkable is that although the same percentage of people changed their direction and
speed, those changes themselves were bigger during the workweek, indicating that in higher
39
densities not more people change direction and speed, but that instead the people who do
change it to a greater degree.
4.2. Comparison of times
There were big differences in walking behavior between morning, midday and
afternoon. Most people were observed during midday, least in the morning. Almost every
measured parameter was higher during midday than in the morning, including direction
change (percentage and degree), speed decrease (percentage and degree), speed increase
(degree) and all personal space values. Contrary to the theory of number of people being the
main factor influencing all other variables there were almost no other differences between
midday and afternoon, although the number of observed people per 30min differed
significantly.
4.3. Comparison of counter areas
When comparing the three counter areas it shows that there wore more people walking
through counter area 1 than through counter areas 2 and 3 combined. Because of the high
density all personal space variables have higher values in this counter area. Direction change
(percentage and degree) was biggest in counter area 3. This can be explained by its location,
lying within a crossing of three hallways, which might force pedestrians to change their
walking direction. All variables concerning speed were highest in counter area 2. This
includes average walking speed, which was at 2.30m/s much higher than described in the
literature (e.g. Bohannon, 1997; Costa, 2010; Henderson and Lyons, 1972; Vallis and
McFadyen, 2003). Speed decrease (percentage and degree) and speed increase (percentage
and degree) were also higher in this area. It is still unclear if people really walked that much
faster in this area or if there was a technical problem, for example caused by the angle of the
sensor.
40
4.4. Density
Density, measured by the number of observed people within each 30min set and the
frequency of people per square meter, correlated with almost every other variable. Confirming
my hypotheses, people walked faster at higher densities, changed their direction and speed to
greater degrees and violated each others personal space more often. However, a smaller
percentage of people decreased their speed in high density situations and there were no
correlations for the percentage of people increasing their speed and changing direction. This
suggests that although the number of people who have to change their walking behavior when
many others are around is the same or even smaller, the degree to which those people change
their walking behavior is bigger (e.g. Gérin-Lajoie et al., 2006; Gérin-Lajoie et al., 2008;
Konečni et al., 1975).
4.5. Direction change
The average angle to which walking direction was changed correlated with most other
variables. Supporting the second hypothesis, namely that higher density leads to more
direction changes, the angle of direction change increased with increasing density and more
violations of personal space. As mentioned above there was no correlation between density
and the percentage of people who changed direction, indicating that at higher densities the
same number of people change walking direction as at lower densities, but do so to a greater
degree.
People who walked slower changed their direction on average more often and to a
greater degree, which makes sense considering that step length needs to be reduced when not
walking straight, thus automatically reducing speed (Chung and Hahn, 1999; Patla, 1996).
We also found that more people changed their direction when the number of violations
of circular personal space was higher, but not when more violations of personal space in
41
motion (PS2/small – PS6/big) occurred. However, as with density, the angle to which people
changed their direction was bigger at higher values for PS2/small – PS6/big.
4.6. Speed
Average speed correlated with the total number of pedestrians, showing that although
people try to keep their desired walking speed whenever it is possible (Helbing et al., 2001;
Ralston, 1958) they still have to change it depending on how many others there are.
Nevertheless there was no correlation between walking speed and personal space intrusions.
Mean walking speed in this sample was with 1,71m/s slightly higher than described in
literature (e.g. Bohannon, 1997; Costa, 2010; Henderson and Lyons, 1972; Vallis and
McFadyen, 2003). However, this was mainly due to the very high average speed (2.30m/s) in
counter area 2. It is unclear if this was a measuring error because of a technical problem of the
system or if people really walked faster in this area. In counter areas 1 and 3 average speeds
laid within the described ranges (1.49m/s for counter area 1 and 1.37m/s for counter area 3).
Therefore, correlations were also calculated excluding counter area 2. This analysis showed
different results, namely that average speed and possible personal space intrusions did in fact
depend on each other, confirming hypothesis 1 that walking speed increases with higher
density because of stress (Konečni et al., 1975). There was no situation where density was so
high that people could not move freely and therefore had to decrease their speed.
When controlling for the number of observed people average speed of all three counter
areas and intrusions into personal space also correlated negatively, adding support for
hypothesis 1.
4.7. Personal space
The main goal of this study was to find out more about the personal space of
pedestrians – how they react to intrusions and how they try to avoid them. Therefore two
42
ways of measuring personal space violations were created; one being circular personal space
around each person, the other one being possible personal space collisions calculated by the
paths people were going, which was measured in six different sizes. In this study those two
approaches brought very similar results. Circular personal space correlated with every other
variable. The six personal space in motion values correlated with every variable except for the
percentage of people changing direction. However, they only correlated with the average
walking speed if either the very high average speed in counter area 2 was excluded or the
analysis was controlled for the number of observed people. As already mentioned, more
violations of personal space occurred when there were more people and at lower walking
speed. Also changes of speed and walking direction were bigger when personal space was
violated more often. However the percentage of people changing speed was lower at the same
time and the percentage of people changing direction did not correlate with personal space.
This shows that when there are more violations of personal space less people change speed
(Konečni et al., 1975) and the same number of people changes direction (e.g. Gérin-Lajoie et
al., 2006; Gérin-Lajoie et al., 2008), but they do so to a greater degree.
Although these results are very interesting and help to increase knowledge about
human walking behavior, it is unfortunate that there were no bigger differences between the
individual personal space values, especially the six values of personal space in motion. Since
this was the first time attempting a calculation like this and there was not much literature
available it is possible that the distances to the sides and to the front that were chosen were
too big to show a significant effect when making them bigger. Also a maximum of three
seconds of planning one’s walking path ahead was used in this study, but it is likely that
people planned their paths further ahead, so that adaptations of walking behavior were the
same for every distance within those three seconds. It would be very interesting to conduct
follow up studies with more extreme differences for those values, to maybe find the point at
43
which pedestrians start changing their paths according to density, and the point at which they
stop doing so.
4.8. CCB Analyser – future implications
The Viennese company Yellowfish GmbH developed the CCB Analyser over the last
years. It is used to monitor crowded public places like train stations or shopping malls,
without invading people’s privacy. The CCB Analyser only shows the movement of crowds,
thus making it impossible to identify individual people or to follow somebody’s path through
the observed area. Therefore it can be used almost everywhere without violating privacy laws.
One part of the present study was to help creating the tools to measure personal space.
Since there has never been a similar approach to this and there is almost no literature about
the personal space of pedestrians it is a very new way to study this topic. As the results show
it still needs some improvement and experimentation, especially regarding the size of personal
space and the time people plan their paths ahead. Also, there might be many other things that
can be measured with CCB Analyser that have not been considered yet, which makes it a
great tool for scientific study about pedestrian behavior in general.
44
V. REFERENCES
Baldassare, M. and Feller, S. (1975). Cultural Variations in Personal Space. Ethos, 3: 481–
503.
Bohannon, R.W. (1997). Comfortable and maximum walking speed of adults aged 20-79
years: reference values and determinants. Age and ageing, 26: 15-19
Buchwalder, T. and Huber-Eicher, B. (2004). Effect of increased floor space on aggressive
behaviour in male turkeys (Meleagris gallopavo). Applied Animal Behaviour Science
89: 207–214
Cheyne, J.A. and Efran, M.G. (1972). The effect of spatial and interpersonal variables on the
invasion of group controlled territories. Sociometry Vol.35, No.3: 477-489
Chung, S. and Hahn, J.K. (1999). Animation of human walking in virtual environments.
Computer Animation, 1999. Proceedings
Costa, M. (2010). Interpersonal distances in group walking. J Nonverbal Behav 34: 15-26
Evans, G.W. and Wener, R.E. (2007). Crowding and personal space invasion on the train:
Please don’t make me sit in the middle. Journal of Environmental Psychology 27: 90–
94
Fajen, B.R. and Warren, W.H. (2003). Behavioral dynamics of steering, obstacle avoidance,
and route selection. Journal of Experimental Psychology: Human Perception and
Performance, Vol.29, No.2: 343-362
Ganem, J. (1998). A behavioral demonstration of Fermat’s principle. The Physics Teacher
Vol.36: 76-78
Gérin-Lajoie, M., Richards, C.L., Fung, J. and McFadyen, B.J. (2008). Characteristics of
personal space during obstacle circumvention in physical and virtual environments.
Gait & posture 27: 239-247
45
Gérin-Lajoie, M., Richards, C.L. and McFadyen, B.J. (2006). The circumvention of obstacles
during walking in different environmental contexts: A comparison between older and
younger adults. Gait & posture 24: 364-369
Gérin-Lajoie, M., Ronsky, J.L., Loitz-Ramage, B., Robu, I., Richards, C.L. and McFadyen,
B.J. (2007). Navigational strategies during fast walking: A comparison between
trained athletes and non-athletes. Gait & Posture 26: 539-545
Hall, E.T. (1966). The hidden dimension. Anchor Books, Doubleday
Hayduk, L.A. (1981). The shape of personal space: An experimental investigation. Canadian
Journal of Behavioural Science, 13: 87-93
Hayduk, L.A. and Mainprize, S. (1980). Personal Space of the Blind. Social Psychology
Quarterly Vol.43, No.2: 216-223
Helbing, D., Molnár, P., Farkas, I.J., and Bolay, K. (2001). Self-organizing pedestrian
movement. Environment and Planning B: Planning and Design 28: 361-383
Henderson, L.F. and Lyons, D.J. (1972). Sexual differences in human crowd motion. Nature
Vol. 240: 353-355
Ickinger, W., Morris, S. (2001). Psychological characteristics and interpersonal distance.
www.sharktown.com/proxemics/intro.html
Johansson, A. (2009). Constant-net-time headway as key mechanism behind pedestrian flow
dynamics. Physical Review E, Vol.80, 026120
Juarbe-Díaz, S.V. (1999). Feline Aggression. Waltham Feline Medicine Symposium. TNAVC
1999
Karlin, R.A., Epstein, Y.M. and Aiello, J.R. (1978). A setting-specific analysis of crowding.
In Baum, A. and Epstein, Y.M. Human response to crowding. Oxford, England:
Lawrence Erlbaum
46
Konečni, V.J., Libuser, L., Morton, H. and Ebbesen, E.B. (1975). Effects of a violation of
personal space on escape and helping responses. Journal of Experimental Social
Psychology 11: 288-299
Middlemist, R.D., Knowles, E.S. and Matter, C.F. (1976). Personal space invasions in the
lavatory: Suggestive evidence for arousal. Journal of Personality and Social
Psychology Vol.33, No.5: 541-546
Moussaïd, M., Helbing, D., and Theraulaz, G. (2011). How simple rules determine pedestrian
behaviour and crowd disasters. PNAS Vol.108, No.17: 6884-6888
Newman, R.C. and Pollack, D. (1973). Proxemics in Deviant Adolescents. Journal of
Consulting and Clinical Psychology Vol.40, No.1: 6-8
Patla, A.E. (1996). Understanding the roles of vision in the control of human locomotion. Gait
& Posture 5: 54-69
Ralston, H.J. (1958). Energy-speed relation and optimal speed during level walking.
Internationale Zeitschrift für angewandte Physiologie, einschließlich
Arbeitsphysiologie, Bd. 17: 277-283
Renner, M.J. and Kelly A.L. (2006). Behavioral Decisions for Managing Social Distance and
Aggression in Captive Polar Bears (Ursus maritimus). Journal of Applied Animal
Welfare Science 9(3): 233-239
Robson, S.K.A. (2008). Scenes from a restaurant: Privacy regulation in stressful situations.
Journal of Environmental Psychology 28: 373-378
Strube, M.J. and Werner, C. (1984). Personal space claims as a function of interpersonal
threat: The mediating role of need for control. J Nonverbal Behav 8(3): 195-209
Taylor, R. (1988). Human territorial functioning. Cambridge University Press
Thompson, D.E., Aiello, J.R. and Epstein, Y.M. (1979). Interpersonal distance preferences. J
Nonverbal Behav 4(2): 113-118
47
Troko, J., Myles, P., Gibson, J., Hashim, A., Enstone, J., Kingdon, S., Packham, C., Amin, S.,
Hayward, A., and Van-Tam, J.N. (2011). Is public transport a risk factor for acute
respiratory infection? BMC Infectious Diseases 2011, 11:16
Vallis, L.A. and McFadyen, B.J. (2003). Locomotor adjustments for circumvention of an
obstacle in the travel path. Exp Brain Res 152: 409-414
48
ZUSAMMENFASSUNG
Menschen haben ein natürliches Bedürfnis, bestimmte physische Abstände zueinander
einzuhalten. Diese Distanzen nennt man Personal Space (oder Personal Distance). Edward T.
Hall (1966) beschrieb sie als eine Distanz von 45 bis 120cm, die zwischen zwei Menschen
eingehalten wird. Sie können einander berühren, wenn sie die Arme ausstrecken und sehen
einander klar, wenn auch nicht vollständig. Dieser Abstand wird immer zu Fremden gehalten,
jedoch können bekannte Personen je nach der Beziehung zwischen beiden näher kommen.
Wird der Personal Space ohne Zustimmung betreten kommt es zu physiologischen
Reaktionen wie erhöhtem Puls, Schwitzen und erhöhtem Blutdruck (Middlemist and
Knowles, 1976). Wir fühlen uns einfach unwohl wenn andere zu nahe kommen.
Personal Space wurde hauptsächlich für stehende und sitzende Testpersonen
untersucht (z.B. Newman and Pollack, 1973; Thompson et al., 1979; Hayduk and Mainprize,
1980; Hayduk, 1981; Strube and Werner, 1984; Evans and Wener, 2007; Robson, 2008), nicht
jedoch für gehende. Gérin-Lajoie und seine Kollegen (Gérin-Lajoie et al., 2006; Gérin-Lajoie
et al., 2008) beschrieben eine Minimaldistanz, die Fußgänger zu stationären und sich
bewegenden Hindernissen halten. Diese wurden als Grundlage für die vorliegende Studie
verwendet. Mit Hilfe eines neuen Systems namens CCB Analyser wurde das Verhalten von
Fußgängern in einem österreichischen Einkaufszentrum untersucht. Dabei wurden Daten zur
Anzahl und Frequenz der Menschen, Durchschnittsgeschwindigkeit, Änderungen von
Geschwindigkeit und Anzahl dieser Änderungen, Änderungen der Gehrichtung und Anzahl
dieser Änderungen, und zwei verschiedene Ansätze zu Messung von Personal Space
aufgenommen. Eine davon beinhaltete die Messung des Personal Space in Kreisen um
Einzelaufnahmen von Personen, die zweite eine Messung der Personal Space für Fußgänger,
die ihre Wege vorausplanen.
49
Die getestete Hypothese lautete, dass hohe Dichte und geringe Distanzen das
Gehverhalten beeinflussen, indem die Geschwindigkeit aufgrund von Stress ansteigt (Konečni
et al., 1975) und Gehgeschwindigkeit und Gehrichtung öfter geändert werden müssen. Die
Ergebnisse der vorliegenden Studie zeigen, dass alle gemessenen Variablen stark voneinander
abhängen. Wir konnten die Hypothese, dass Menschen schneller gehen wenn ihnen andere zu
nahe kommen, zumindest teilweise bestätigen. Fußgänger ändern ihre Geschwindigkeit und
Gehrichtung mehr bei hoher Dichte, jedoch bleibt der Prozentsatz der Menschen, die ihr
Gehverhalten ändern, gleich oder sinkt sogar.
Diese Ergebnisse sollen einen ersten Einblick in das Verhältnis zwischen
menschlichem Gehverhalten und Personal Space bieten, jedoch muss zu diesem Thema noch
viel Forschung betrieben werden.
50
ABSTRACT
Humans have a natural desire to keep a certain physical distance from other humans.
This distance is called personal space (or personal distance). Edward T. Hall (1966) describes
it as a distance of 45 to 120cm kept from each other, a which people can touch if they extend
their arms and see each other clearly, but not as a whole. Humans always try to keep this
minimum distance to strangers, but might let familiar people closer, depending on their
relationship. If personal space is invaded without consent it comes to physical reactions such
as increased heart rate, sweating and increased blood pressure (Middlemist and Knowles,
1976). We simply feel uncomfortable when others come too close.
Personal space is well described for standing and seated test subjects (e.g. Newman
and Pollack, 1973; Thompson et al., 1979; Hayduk and Mainprize, 1980; Hayduk, 1981;
Strube and Werner, 1984; Evans and Wener, 2007; Robson, 2008), but not for walking
people. Gérin-Lajoie and his colleagues (Gérin-Lajoie et al., 2006; Gérin-Lajoie et al., 2008)
described minimum distances that pedestrians keep from stationary and moving obstacles,
which were used as a basis for this study. Using a newly developed system called CCB
Analyser the walking patterns of pedestrians in an Austrian shopping center were recorded.
Data included number and frequency of people, average speed, speed changes and number of
speed changes, direction changes and number of direction changes, and two different
measures for personal space, one being personal space in circles around stationary recording
frames and the other being personal space for pedestrians that plan their paths ahead.
The tested hypothesis was that high density and low interpersonal distance leads to a
change of walking behavior – increasing walking speed because of stress (Konečni et al.,
1975) and making people change their speed and directions when walking. The results of the
present study show that all measured variables seem to highly depend on each other. We
could at least partly confirm the hypothesis of people walking faster when personal space is
51
invaded. People changed their walking speed and direction to a higher degree at high
densities, however the percentage of people changing their walking behavior was the same or
even smaller.
These results offer a first insight into the relationship of human walking behavior and
personal space, but much more research needs to be done on this topic.
52
Curriculum vitae
Personal information
Name Anna Frohnwieser
Date of birth 27.04.1987
Nationality Austria
Address Rueppgasse 38/21
1020 Vienna
Austria
Mobile +4369911801386
E-Mail annafrohnwieser@hotmail.com
Education
1993 – 1997 Volksschule St. Margarethen/Bgld.
1997 – 2000 BG/BRG/BORG Eisenstadt
2000 – 2005 BG und BRG Neusiedl am See
June, 7th 2005 Matura
October 2005 – July 2007 Studies of Biology at the University of Vienna
Since July 2007 Studies of Anthropology and Zoology at the University of
Vienna
53
