ARTICLE IN PRESS

Journal of Environmental Psychology 27 (2007) 90–94

www.elsevier.com/locate/jep

Crowding and personal space invasion on the train: Please don’t make
me sit in the middle
Gary W. Evansa,b,, Richard E. Wenerc
a
Department of Design and Environmental Analysis, Cornell University, Ithaca, NY 14853-4401, USA
b
Department of Human Development, Cornell University, Ithaca, NY 14853–4401, USA
c
Polytechnic University, USA
Available online 23 January 2007

Abstract

Mass transit users frequently experience crowding during their commutes. In this study of 139 urban passenger train commuters during
rush hour, we found that the density of the train car was inconsequential for multiple indices (self-report, salivary cortisol, performance
aftereffects) of stress whereas the immediate seating density proximate to the passenger significantly affected all three indices. When
people had to sit close to other passengers, they experienced adverse reactions. These results are consistent with prior work indicating
that individual spacing among persons that leads to personal space invasions is a more salient environmental condition than density per
se. The findings also have implications for the design of mass transit vehicles.
r 2006 Elsevier Ltd. All rights reserved.

Keywords: Crowding; Personal space; Stress; Commuting

1. Introduction increased, rather than occupying available seats, many

Crowding on trains is well known to commuters who use
mass transit for rush hour trips. A recent report to the city
of London attributed a variety of health and safety
problems on trains to crowding (Turner, Corbett, O’Hara,
& White, 2005). Transit authorities in the New York
Metropolitan Area have noted widespread dissatisfaction
about crowding among passengers using new rail cars, even
though each car actually holds a smaller number of
passengers, and in some cases, even provides wider seats
(Mc Geeham, 2005). What the designers of these new cars
may have overlooked, however, is the role of personal
space invasion on trains. Mc Geeham’s article also noted
the large number of vacant middle seats on trains, even
when trains were filled. This observation fits well with a
more systematic study of proxemic behavior on subway
cars in New York city conducted three decades earlier by
Fried and DeFazio (1974). Among many fascinating
observations, the authors noted that as car density
Corresponding author. Tel.: +1 607 255 2144; fax: +1 607 255 0305.
E-mail address: gwel@cornell.edu (G.W. Evans).
passengers opted instead to stand in places that maximized
physical distance from other passengers. When extreme
levels of density precluded this option, all seats were then
filled and various nonverbal indices of avoidance (e.g.,
minimal eye contact, bodies turned towards the sides of the
car and away from people, arms crossed) were markedly
elevated. In this paper, we explore the role of personal
space invasion and crowding among train commuters
during rush hour in New York City.
Conceptual models have related crowding in situations
like these to difficulties in privacy regulation. We regulate
privacy by adjusting social interaction to desired levels
through an intricate system of verbal, nonverbal, and
physical processes (Altman, 1975). For example people use
culturally accepted norms of interpersonal distance—
personal space—to help regulate privacy. Crowding occurs
when the regulation of social interaction is unsuccessful
and our desires for social interaction are exceeded by the
actual amount of social interaction experienced. Perceived
crowding is related, but not identical, to the density
of a situation, which is typically operationalized as the
relation between numbers of people and unit of space

0272-4944/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jenvp.2006.10.002
 ARTICLE IN PRESS
G.W. Evans, R.E. Wener / Journal of Environmental Psychology 27 (2007) 90–94
(e.g., people per room) (Baum & Paulus, 1987; Evans,
2001; Stokols, 1972).
It has long been recognized, however, that density does
not adequately capture the experience of individuals in a
given space since the distribution of both people (Day &
Day, 1973; Knowles, 1978) and of physical objects in
concert with physical layout of the space (Evans, 1979) can
directly influence the degree of physical proximity to
others. Intrusions into personal space then may be a
fundamental cause of crowding. As density grows, the
probability elevates that one’s personal space will be
intruded upon, creating discomfort in and of itself (Aiello,
1987; Kaya & Erkip, 1999). Higher density also increases
the likelihood of physical and social contact and accom-
panying undesired and unpredictable interactions, which
have also been identified as important facets in the
experience of crowding (Baum & Valins, 1977; Saegert,
1978). We also know from laboratory (Evans & Howard,
1972; McBride, King, & James, 1965) and field studies
(Middlemist, Knowles, & Matter, 1976), that involuntary
physical proximity to another individual elevates physio-
logical stress. In this paper we examine a naturalistic situ-
ation that is often experienced as crowded, commuting on
public transit during rush hour. We address the question, is
the adverse impact of increasing density in a naturally
crowded situation, primarily a function of personal space
intrusion? We believe it is the actual experience of overly
close physical proximity to another individual that under-
lies the experience of crowded commuting.
Both Worchel and Teddlie (1976) and Sundstrom,
Busby, and Asmus (1975) directly tested the hypothesis
that a critical component of the experience of crowding was
personal space intrusion. These studies varied density and
interpersonal distance in a factorial design in the labora-
tory, and in each case personal space invasion, not overall
density, was the key factor for perceived crowding, task
performance, and nonverbal indices of social withdrawal.
Under more naturalistic conditions, personal space inva-
sions but not density largely influenced nonverbal indica-
tors of discomfort between strangers (Harris, Luginbuhl, &
Fishbein, 1978) and at the office, distances between
workstations had a greater influence on worker’s social
withdrawal behaviors compared to the effects of office
density (Oldham & Fried, 1987).
We examined the contributions of train density and seat
proximity to multimethodological indices of stress among
passenger train commuters. These results are not only
interesting theoretically but as indicated at the onset, have
direct implications for the design of mass transit systems.
One of the primary reasons people give for commuting to
work by automobile compared to public transit is
enhanced privacy provided by cars (Dockendorf, Levinson,
Fichter, Haghani, & Prestrud, 2001; Ibrahim, 2003; Joire-
man, Lange, Kuhlman, Vugt, & Shelley, 1998; Petkewich,
2005). If we can better understand what aspects of the
commuting experience contribute to privacy, we may be
able to design more appealing public transit options.
91
2. Method
2.1. Participants
One hundred and thirty-nine adult commuters (54%
male) who had been commuting to work by passenger
train from the same geographic area of New Jersey into
Manhattan, New York City were recruited into the study.
Each participant was provided a free monthly rail pass for
their participation. Participants had been on the same
commuting route for an average of 82 months with a 12-
month minimum. The average duration of the commute to
work was 83 min. The median income level exceeded
$95,000 and 83% had a college degree.
2.2. Procedure
Salivary cortisol was collected with a Salivette in the
train station at the end of the participant’s morning trip to
work and at the same time of day on the following weekend
at the respondent’s home. Samples were placed under
20 1C until assay. Cortisol was analysed by a time-
resolved immunoassay with fluorescence detection (Dres-
sendorfer, Kirschbaum, Rohde, Stahl, & Strasburger,
1992). The difference between resting and on-task, or at
work, neuroendocrine hormones has been shown to be a
reliable and valid marker of stress both in the laboratory
and the field (Frankenhaeuser & Johansson, 1986; Kirsch-
baum & Hellhammer, 1989; Lundberg, 1984; Lundberg,
2000).
Motivation was measured near the end of the morning
commute by persistence on a proofreading task (percent
errors detected). This measure has been widely used to
assess aftereffects from stressor exposure both in the
laboratory and the field (Cohen, 1980; Glass & Singer,
1972). Mood was assessed during the commute to work by
two, five point semantic differential scales (carefree–bur-
dened; contented–frustrated), a ¼ :74.
Public transit crowding was assessed at the distal and
proximal level. Distally, we calculated the density of the
passenger train car by dividing the number of people in
the car at the time of the passenger’s entry by the number
of seats on the car. Seat density, a more proximal index
of crowding, was calculated by dividing the number of
people sitting in the same immediate row the passenger
was seated in by the number of total seats in the row.
Fig. 1 provides a diagram of a typical seating pattern on
the train.
Fig. 1. Train car seating configurations.
 ARTICLE IN PRESS
92 G.W. Evans, R.E. Wener / Journal of Environmental Psychology 27 (2007) 90–94
Table 1
Descriptive statistics and zero-order correlations
Variable Mean
1. Car density (people/number of seats in car) .37
2. Seat density (people/number of seats in row) .44
3. Cortisol (log commuting–log at home baseline) .13
4. Proofreading (% errors detected) 55
5. Mood (2–10) 5.47
6. Income (1o$25,000–94$95,000) 7.33
po.05 (one-tailed).
3. Results
Car and seat density were examined as predictors of
stress with multiple regression with income as a statistical
control. Means, standard deviations and zero-order
correlations are shown in Table 1. As can be seen in the
table seat density but not car density is correlated with
some of the stress measures.
All of the effects of car density remain nonsignificant
with controls for income, cortisol, b ¼ :43 (.36), ns;
proofreading, b ¼ :09 (.06), ns; and mood, b ¼ :00
(.51), ns. We statistically controlled for income given the
typical covariation between income and house location,
commuting route, as well as overall levels of life stress. The
standard error for each respective raw beta weight is in
parentheses. For seat density all three stress indices are
significant for crowding immediately around the partici-
pant with income as a statistical control, cortisol b ¼ :70
(.34), po.02, proofreading b ¼ :09 (.06), po.05 and
mood b ¼ :78 (.49), po.05.
4. Discussion
The results of this analysis of crowding on public transit
during the morning rush hour are largely in accord with
prior research on crowding both in the laboratory and
when observed under naturalistic conditions. We find that
a more proximal index of density is correlated with
multiple indices of stress wherein a more distal index of
density is not. Specifically, the ratio of the number of
people on a passenger car to the total number of seats (car
density) during rush hour is not related to stress whereas
the ratio of the number of people to seats in the immediate
area of the passenger is. In some respects, these findings
support Altman’s (1975) observation that indices of density
that reflect the individual’s actual experience of density or
the close presence of other people will be more salient in
accounting for human behavior than more distal indices of
density that simply represent the average amount of space
available to people. More recent investigations of the
measurement of crowding are in accord with this conclu-
sion (Evans, 2001).
Both Day and Day (1973) and Knowles (1978)
recognized early on that density as a measure of people
per area does not fully capture the experience of crowding
SD 2 3 4 5 6
.29 .13 .11 .13 .00 .06
.28 .19* .12 .16* .03
1.11 .09 .12 .06
19 .17* .12
1.58 .05
2.22
since people can distribute themselves in many different
ways in a given space. Furthermore, floor plan layout
plus the positioning of furniture and other physical barriers
can alter the individuals actual experience of crowding
(Evans, 1979; Stokols, 1972b). Our results suggest the
immediate close presence of other passengers is probably
more salient for the experience of crowding on public
transit than the total number of people on the car. If
anything, our results probably underestimate the impact
of personal space intrusions on stress since we did not
interact with the passengers and thus cannot determine
whether the person they were sitting next to was a family
member, friend or a stranger. Considerable research
indicates that invasions of space by strangers produce
more stress than those occurring among people with
positive interpersonal relationships (Aiello, 1987; Evans
& Howard, 1973).
As noted earlier, journalistic reports suggest that people
find middle seats on passenger trains especially aversive,
leading to problems for transit systems in designing cars to
comfortably accommodate expected capacities (Mc Geeham,
2005). A plausible explanation for this commuter response
is personal space intrusion. One of the more salient
characteristics of higher density settings is the enhanced
probability that one will be forced into a situation where
one may find oneself standing or sitting too close to
another person (Aiello, 1987; Kaya & Erkip, 1999;
Worchel & Teddlie, 1976). This is particularly egregious
when: (a) one has little control over the situation and
(b) this happens with strangers. Both of these character-
istics are commonly met during rush hour commuting on
public transit. The finding that sitting close to someone as
opposed to being on a train car with a larger number of
people is especially stressful is also consistent with
Altman’s (1975) privacy model. Personal space may act
as a boundary control mechanism that enables us to
optimize the degree of social interaction obtained at any
one given point in time. When this mechanism becomes
restricted or ineffective, such as when a stranger sits down
next to you on a train, greater social interaction than
desired will frequently occur. Seating in a middle as
opposed to an end seat doubles the likelihood and of
spatial intrusions and reduces opportunities for adjusting
to the situation (e.g., turning away from the proximate
seatmate).
 ARTICLE IN PRESS
G.W. Evans, R.E. Wener / Journal of Environmental Psychology 27 (2007) 90–94
Our interpretation of the present findings suggests a
number of possible design interventions. Public transit
designers could provide pairs of proximate seats on public
transit vehicles, instead of three across seating, or include
territorial props (e.g., arm rest, small table) in between
seats to facilitate the user’s ability to regulate social
interaction (Baum, Riess, & O’Hara, 1974; Dumur,
Barnard, & Boy, 2004; Fisher & Byrne, 1975). Larger cars
that hold more people could be incorporated to help
compensate for the loss of seat space (and revenue) if our
finding that overall density of the car is relatively
unimportant in people’s commuting experience is robust.
The correlation between seat and car density (r ¼ :13)
was lower than expected. One would think that as
passenger train cars filled up with people, individuals
would then occupy more seats, including those nearby
strangers. However as indicated in Fried and DeFazio’s
(1974) ethnographic study of subway riders in New York
City, many passengers instead opted to stand by themselves
rather than sit next to a stranger. In hindsight, we should
have counted the number of standees in addition to noting
seating positioning. Anecdotally, one of the authors of the
present paper is a regular train commuter to New York
City and has frequently witnessed standees when seats were
available on commuter trains.
Another limitation of the present study in addition to the
cross-sectional design, is the small magnitude of the effects
of seat density on stress outcomes (see Table 1). On the
other hand although our effect sizes are small, the
magnitude of changes in stress as a function of seat density
are not trivial. For example, a one standard deviation
increase in seat density after controlling for income elevates
the log of cortisol by over 2000% and diminishes task
performance and mood by 6% and 4%, respectively.
Although seat density effects are consistently, statistically
significant and car density effects are consistently not
significant, a more stringent test of our hypothesis would
be that seat density effects are also significantly greater
than car density effects. This is true only for mood. Where
another person sits relative to your position in the public
transit environment appears to be salient to the commuting
experience. At the same time, how many people are present
relative to the total number of seats (car density), appears
to be less important to commuting stress. More fine tuned
indices of density that more fully capture the actual
experience of physical proximity to other people may
prove more potent in predicting health and behavioral
outcomes of crowding. How easily and how frequently
one’s personal space is intruded upon may be one of the
key underlying processes that underlie the experience of
crowding.
References
Aiello, J. R. (1987). Human spatial behavior. In D. Stokols, & I. Altman
(Eds.), Handbook of environmental psychology (pp. 359–504). New
York: Wiley.
93
Altman, I. (1975). The environment and social behavior. Monterey, CA:
Brooks-Cole.
Baum, A., & Paulus, P. B. (1987). Crowding. Handbook of environmental
psychology, 533–570.
Baum, A., Riess, M., & O’Hara, J. (1974). Architectural variants of
reactions to spatial intrusion. Environment and Behavior, 6, 91–100.
Baum, A., & Valins, S. (1977). Architecture and social behavior:
Psychological studies of social density. Hillsdale, NJ: Erlbaum.
Cohen, S. (1980). Aftereffects of stress on human performance and social
behavior: A review of research and theory. Psychological Bulletin, 88,
82–108.
Day, A. T., & Day, L. H. (1973). Cross-national comparison of
population density. Science, 181, 1016–1023.
Dockendorf, J., Levinson, H., Fichter, D., Haghani, A., & Prestrud, C.
(2001). Bus transportation: A look forward. Washington, DC:
Transportation Research Board.
Dressendorfer, R. A., Kirschbaum, C., Rohde, W., Stahl, F., & Strasburger,
C. J. (1992). Synthesis of cortisol–biotin conjugate and evaluation as a
tracer in an immunoassay for salivary cortisol measurement. Journal of
Steroid Biochemistry and Molecular Biology, 43, 683–692.
Dumur, E., Barnard, Y., & Boy, G. (2004). Designing for comfort. In D.
de Waard, K. A. Brookhuis, & C. B. Weikert (Eds.), Human factors in
design (pp. 111–127). Maastricht, The Netherlands: Shaker.
Evans, G. W. (1979). Design implications of spatial research. In J. R.
Aiello, & A. Baum (Eds.), Residential crowding and design
(pp. 197–216). New York: Plenum.
Evans, G. W. (2001). Environmental stress and health. In A. Baum, T.
Revenson, & J. E. Singer (Eds.), Handbook of health psychology
(pp. 365–385). Hillsdale, NJ: Erlbaum.
Evans, G. W., & Howard, R. B. (1972). A methodological investigation of
personal space. In W. J. Mitchell (Ed.), Environmental design, research,
and practice: Proceedings of the third environmental design research
association conference (pp. 221–228). Los Angeles: University of
California Press.
Evans, G. W., & Howard, R. B. (1973). Personal space. Psychological
Bulletin, 80, 334–344.
Fisher, J. D., & Byrne, D. (1975). Too close for comfort: Sex differences in
response to invasions of personal space. Journal of Personality and
Social Psychology, 32, 15–21.
Frankenhaeuser, M., & Johansson, G. (1986). Stress at work: Psychobio-
logical and psychosocial aspects. International Review of Applied
Psychology, 35, 287–299.
Fried, M., & DeFazio, V. J. (1974). Territoriality and boundary conflicts
in the subway. Psychiatry, 37, 47–59.
Glass, D. C., & Singer, J. E. (1972). Urban stress: Experiments on noise and
social stressors. New York: Academic Press.
Harris, B., Luginbuhl, J. E., & Fishbein, J. E. (1978). Density and personal
space in a field setting. Social Psychology, 41, 350–353.
Ibrahim, M. F. (2003). Car ownership and attitudes towards transport
modes of shopping purposes in Singapore. Transportation, 30,
435–457.
Joireman, J. A., Lange, P. A., Kuhlman, M., Vugt, M. V., & Shelley, G. P.
(1998). An interdependence analysis of commuting decisions. European
Journal of Social Psychology, 27, 441–463.
Kaya, N., & Erkip, F. (1999). Invasions of personal space under the
condition of short-term crowding: A case study on an automated teller
machine. Journal of Environmental Psychology, 19, 183–189.
Kirschbaum, C., & Hellhammer, D. (1989). Salivary cortisol in
psychobiological research. Neuropsychobiology, 22, 150–169.
Knowles, E. S. (1978). The gravity model of crowding: Application
to social physics to the effects of others. In A. Baum, & Y. Epstein
(Eds.), Human responses to crowding (pp. 183–218). Hillsdale, NJ:
Erlbaum.
Lundberg, U. (1984). Human psychobiology in Scandinavia ii. Psycho-
neuroendocrinology: Human stress and coping processes. Scandinavian
Journal of Psychology, 214–226.
Lundberg, U. (2000). Catecholamines. In G. Fink (Ed.), Encyclopedia of
stress, Vol. 1 (pp. 408–413). New York: Academic Press.
 ARTICLE IN PRESS
94 G.W. Evans, R.E. Wener / Journal of Environmental Psychology 27 (2007) 90–94
McBride, G., King, M., & James, J. (1965). Social proximity effects on galvanic
skin response in adult humans. Journal of Psychology, 61, 153–157.
Mc Geeham, P. (2005). For train riders, middle seat isn’t the center of
attraction. New York Times (pp. A-1 and B-5).
Middlemist, R. D., Knowles, E. S., & Matter, C. F. (1976). Personal space
invasions in the lavatory: Suggestive evidence for arousal. Journal of
Personality and Social Psychology, 33, 541–546.
Oldham, G., & Fried, Y. (1987). Employee reactions to workspace
characteristics. Journal of Applied Psychology, 72, 75–80.
Petkewich, R. (2005). Taking the mass out of transit. Environmental
Science and Technology, 39, 61a.
Saegert, S. (1978). High density environments: Their personal and social
consequences. In A. Baum, & Y. Epstein (Eds.), Human response to
crowding, Vol. (pp. 259–282). Hillsdale, NJ: Erlbaum.
Stokols, D. (1972). On the distinction between density and crowding:
Some implications for future research. Psychological Review, 79,
275–277.
Sundstrom, E. B., Busby, P., & Asmus, C. (1975). An experimental study
of crowding: Effects of room size, intrusion, and goal blocking on
nonverbal behavior, self-disclosure, and self-reported stress. Journal of
Personality and Social Psychology, 32, 645–654.
Turner, S., Corbett, E., O’Hara, R., & White, J. (2005). Health and
safety effects of crowding—hazard identification (No. T307).
London: Health and Safety Laboratory, Rail Safety and Standards
Bureau.
Worchel, S., & Teddlie, C. (1976). Factors affecting the experience of
crowding: A two-factor theory. Journal of Personality and Social
Psychology, 34, 30–40.