USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME

LINDA N. NUBANI
The American University in Dubai, P. O. Box: 28282, Dubai, United
Arab Emirates
[email protected]

Abstract. Space syntax provides methods for analyzing spaces using

recent developments in computer programs. This paper reports a
study that was undertaken to investigate the role of space syntax in
identifying geographical patterns of crime in Ypsilanti, Michigan. All
the spaces in the city were analyzed using the Spatialist, a computer
program developed by Georgia Tech. The Spatialist computes the
accessibility level of all the spaces in a spatial system.
Sociodemographic variables such as median income, racial
composition, youth concentration and level of education were
available from the U.S. Census. The crime report was obtained from
the Ypsilanti Police Department and Eastern Michigan University. It
includes data on four types of crime at an address level with the exact
date and time. Both sociodemographic variables and crime data were
merged with the Spatialist map using ArcGIS. The data was analyzed
using SAS, an advanced statistical package. Findings showed strong
relationships between attributes of space and crime locations.

1. Introduction

A considerable body of literature now exists on the impact of spatial

configuration on human behavior. Wayfinding performance, for example,
decreases in buildings characterized with high number of hallway
intersections (Best, 1970) and with more complex relations between choice
points (O’Neill, 1991). On an urban scale, some studies have given some
weight to assertions that curvilinear street networks, T-intersections, and
spatial relationships influence walking and bicycling (Frank and Engelke,
2001; Southworth and Owens, 1993.). Much of these researches, however,
were not able to illuminate the extent to which spatial configuration
influences human behavior. Additionally, it was difficult to derive an
objective methodology that can reliably demonstrate a statistical relationship
between environment and behavior.
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 383

Only recently, some researchers have begun to give more attention to the
properties of spatial configuration and their effect on human behavior using
Space Syntax techniques. Space syntax, a group of theories that examine the
social use of space, was developed in the late 60s by Hillier and Hanson
(1984). Briefly, this theory is based on breaking all the spaces in a plan into
long lines of sight. For example, all the hallways in a building plan will be
flood-filled by long lines of sight. Similarly, all the streets in a city will be
broken into interconnected long lines of sight. According to space syntax,
these lines are known as axial lines.
Gradually and over the years, several space syntax measures were
developed based on the relation between each axial line and all the lines in
the system (e.g. in a building plan or in a city). This research will focus on
two of these measures: Integration and Connectivity. To elaborate on these
two measures, Connectivity gives the number of lines that are directly
connected to a specific line. Integration, on the other hand, is an indicator of
how easily a line is reached from all other lines in the spatial area.
Mathematically speaking, it is the average number of spaces that one needs
to pass through to reach a specific line from all other axial lines in the
system. In other words, integration values suggest the extent to which a
selected space in the system is more integrated (can be easily reached from
other spaces), or more segregated (one has to travel through many spaces in
order to reach that selected space). Two additional measures were also
derived from integration measure: 1) global integration measures the
relationship between a specific line and all the other lines in a system, and 2)
local integration measures the accessibility of the line from lines that are few
steps away as specified by the researcher.
Several computer programs have been developed to compute these
measures. Most of these programs produce two types of output:
alphanumeric data with spatial parameters assigned to each axial line and
graphic data with a map that has colored lines where red indicates the most
integrated or connected line to indigo for the most segregated or least
connected line. Simply put, looking at the resulting map one will enable the
researcher to make inferences about streets, areas or neighborhoods in a city.
Examples of these programs include the Spatialist, developed by Peponis
and Wineman at Georgia Tech; Depthmap, developed by Turner at
University College London; and Omnivista, developed by Dalton and Dalton
at University College London.
As a result of advancement in computer technologies, several studies
have looked at the implications of space syntax measures on human
behavior. For example, we investigated the effects of spatial behaviors and
layout attributes on individual’s perception of psychosocial constructs in
four U.S. federal office settings (Rashid et al., 2005). Using space syntax
techniques, questionnaire surveys and behavioral mappings, we found that
384 LINDA NUBANI

measures like integration and connectivity had significant effects on

individual’s perception of psychosocial constructs in office settings.
On an urban scale, several studies found strong correlations between
space syntax measures and walking behavior. Reid (1999), for example,
found that Deutch neighborhoods with higher mean connectivities and
integration had higher movement rates with a correlation coefficient of
0.829. Similarly, Nubani (2003) confirmed similar results in three cities in
Southeastern Michigan. Correlations between integration and walking were
strong.
The objective of this paper is to use similar computer technologies in
understanding geographical distribution of crime as a behavioral outcome. It
examines the relationship between measures of space syntax and certain
types of crime in order to address the question of ‘opportunity.’ Specifically
the research applies the techniques of space syntax to explore the
characteristics of streets that might contribute to the reduction of
opportunities for criminal acts. In addition to examining street
characteristics, the paper also addresses theories related to crime as set forth
by criminologists.

2.Background Literature

Generally, most of the ‘design for crime prevention’ work has been
grounded in three theories related to crime: the rational offender theory (also
known as rational choice theory), the behavioral geography theory and the
routine activities theory (Taylor, 2002). The rational offender theory
stemmed from the classical school of criminology founded by Cesare
Beccaria. He believed that people have the will to act freely and that crime
is controllable by means of punishment. However, this perspective declined
by mid-twentieth century as the positivist criminology argued that crimes are
the results of genetic, social and psychological factors rather than personal
choice and decision-making (Siegel, 2002). In 1970s, the rational choice
theory supported the thoughts called by the classical school of criminology
and assumed that the benefits of crime influence patterns of offenses. In
other words, criminals are rational actors who weigh the potential costs of
crime and the consequences of their actions (Siegel, 2002). Their decision
process considers both their personal needs such as money and excitement
and situational factors such as the likelihood of being caught and
surveillance.
The behavioral geography theory, on the other hand, considers the fact
that places that are closer to where offenders work or reside are at higher risk
of being burglarized than places that are not within the offenders’ regular
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 385

route. One may infer that this theory suggests that crime rate is linked to
easy accessibility (Taylor, 2002).
The routine activities theory looks at the interaction of three everyday
variables: 1) the availability of attractive targets such as unlocked homes and
attractive valuables; 2) the absence of guardians such as neighbors,
homeowners or police; and 3) the presence of motivated offenders such as
teenagers and unemployed people (Reid, 2002; Siegel, 2002). If targets are
exposed to all three variables, they are at higher risk of being victimized.
It is worth mentioning that the rational offender theory provided the basis
for the situational crime prevention program (Bennett, 1989). It is targeted
at reducing the opportunities to commit burglaries based on the belief that
offenders freely and actively commit crimes as a response to immediate
circumstances and depending on costs and rewards.
Based on the review provided by Bennett (1989), situational measures
operate at three levels. First, at an individual level, situational measures call
for target hardening and installing alarms and surveillance cameras. The
time it takes to overcome such obstacles is perceived as a risk by offenders.
Secondly, at a community level, neighborhood watch programs have been
implemented to involve residents in reporting suspicious activities.
Research to date has not shown whether this had a perceived risk among
offenders or not. Thirdly, at a physical environment level, situational
measures are based on Jacobs and Newman’s concepts of controlling
pedestrian and traffic flows, territoriality and natural surveillance. Jacobs
believed that through the occupation and use of space, residents come to
consider a particular space as theirs and exert control over it (Jacobs, 1961).
To a certain degree, it can be deduced from the afore-mentioned theories
and prevention programs that offenders share four general concerns: how
quickly it takes to get to the target, how quickly it takes to run away, how
much value the target possibly has, and, how likely the offender is to be
caught while committing the crime or leaving the scene (Taylor, 2002;
Rengert, 1980).
Previous literature has also shown that the three basic elements necessary
for someone to commit a crime are ability, opportunity and motive (Stollard,
1991). Thus, if it could be shown using space syntax techniques which
streets offer the opportunity to commit a crime, then it becomes easier for
police to know which streets to increase patrolling

3. Space Syntax and Crime

Building on the idea that neighborhood layouts provide opportunities and

access to commit a crime, Shu and Huang (2003) studied the influence of
spatial configuration on the distribution of burglary in 121 residential
386 LINDA NUBANI

neighborhoods. In the first part of their analysis, they controlled for social
factors by looking at three districts in Northern Taiwan inhabited by
different social classes. Police crime data was gathered for an 8 month
period; there were total number of 241 crime incidents. Through
correlational analyses, a strong connection was found between global
integration and burglary rates in low-income neighborhoods. Further
findings indicated that there were correlations between local integration and
global integration and burglary rates in middle-income neighborhoods. The
authors proposed that globally and locally integrated middle-income
neighborhoods are safer than segregated ones. In addition, the authors found
no correlation between global or local integration and burglary rates in high-
income neighborhoods. This is possibly explained by the fact that “target
hardening” features are more common within high income neighborhoods.
Similar to previous work by Shu and Huang, Jones & Fanek (1997)
looked at the effect of spatial configuration on crime in Austin, Texas. They
selected four pairs of tracts in which each pair had similar income, poverty
rates, population and racial composition. Using Axman software, developed
at University College London, their findings showed that pairs with higher
integration values were associated with lower crime rates. The authors
explained that more connected streets will attract higher pedestrian
movement, and thus more eyes on the street.
As a result of promising findings using Space Syntax for identifying the
spatial distribution of crime, Gosnells, a city in Western Australia consulted
the Space Syntax laboratory at University College London and Murdoch
University to identify the spatial distribution of crime (Australia’s National
Government Newspaper, 2003). The Space Syntax Lab compared the
movement of pedestrians and vehicles to crime statistics and space syntax
measures. The results were consistent with previous findings and showed a
strong link between spatial configuration and burglary and theft.
Farooq (1999) looked at crime in Metro Atlanta in his doctoral
dissertation. He investigated spatial and sociodemographic measures in
different types housing settings using similar computer programs. Contrary
to previous research, his findings showed that in private rental housing and
public housing, crimes against property and persons were higher in
integrated areas. The author suggests that this is explained by the fact that
these buildings were located on vehicular routes that offered an easy escape
to offenders.

4. Types of Crime and Description of Case Study

Generally, different types of crime have been found to be associated with

different types of land use and social characteristics (Dunn, 1980). Personal
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 387

attack crimes, for example, occur more often in lower class neighborhoods,
while property crimes occur more often in neighborhoods that are accessible
or close to other land uses, or in neighborhoods with higher percentages of
underemployed or single residents. Arsons, robberies and burglaries share
monetary gain objectives and are more likely to occur in middle- and high-
class neighborhoods (Rengert, 1980). For these reasons, we excluded non-
residential neighborhoods from this study. We also excluded organized
crimes or crimes that involve acquaintances or for the purpose of revenge
such as assaults and murder. Specifically, we focused on four stranger-to-
stranger types of crime. These are larceny, motor vehicle theft, breaking and
entering, and robbery.
According to the FBI Uniform Report (1998), larceny, motor vehicle
theft and breaking and entering are considered property crimes where the
object of the offense is the taking of property without any threat involved.
More precisely, larceny is taking away property from the possession of
another. Purse-snatching and shoplifting are good examples of larceny.
Motor vehicle theft is the stealing of a truck, automobile, motorcycles, and
any other vehicle. Breaking and entering is defined as the unlawful entry
into a property without putting people under threat (Hill, 1995). Robbery on
the other hand is a violent crime that involves putting victims under threat.
It includes taking anything of value from persons (FBI Uniform Report,
1998).
In this study, we looked at Ypsilanti, a city located within the
Metropolitan Detroit area of Michigan. With a population of approximately
22,362, 1273 crime incidents were reported in year 2003. Crime types in
this figure include larceny, breaking and entering, robbery and motor vehicle
theft. According to FBI Crime Reports, the crime level in Ypsilanti is worse
than the national average particularly for burglaries, robberies, and thefts
(FBI Crime Reports, 2002). The crime report was obtained from the
Ypsilanti Police Department and Eastern Michigan University. It includes
data on the four types of crime at an address level with the exact date and
time.

5. The Axial Map Analysis

A street map of Ypsilanti was imported into the Spatialist program. All the
streets were then broken into long lines of sight that if two people stand at
each end of the line, they should be able to see each other. These lines are
also known as axial lines. Ypsilanti comprised of an average of 2000 axial
lines. Analyzing the relationship among these lines is impossible to
calculate manually. The Spatialist assigned an ID to each axial line and
appended three space syntax measures to each line. These measures were
388 LINDA NUBANI

connectivity, local integration and global integration. The Spatialist also

produced a colored graphic representation of these values on the map of
Ypsilanti (see Figure 1). The colors range from red indicating highly
accessible routes (high values) to indigo indicating the least accessible routes
(low values). Accessibility is explained in terms of the average number of
turns one needs to make to get to any part of a street from any point in the
city. As explained in the next section, a glance at this map tells the
researcher which of the streets offered escape route to criminals.

Figure 1. The Spatialist output of the Ypsilanti Axial Map.

Since the unit of analysis is the axial line (or the street space), it was
necessary to append sociodemographic data along with crime data to each
line. Therefore, a street map of Ypsilanti was prepared showing 30 block
groups using ArcGIS. Data on population density, youth concentration,
level of education, percentage of owners, age distribution and racial
composition were available from U.S. Census and were appended to each
block group in Ypsilanti. The report on crime at an address level was semi-
manually entered into the same database. Moreover, the original axial map
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 389

that was prepared using Spatialist was later converted into an appropriate
format and was given accurate geographic coordinates for Ypsilanti,
allowing us to match the Spatialist axial map with the ArcGIS Ypsilanti road
map (Figure 2), and to merge our two databases.

Figure 2: Crime locations plotted on Ypsilanti Axial Map. Line weight indicates level of
connectivity where thick lines represent highly connected spaces and thin lines represents
spaces with low number of connections

6. Results and Analysis

The GENMODE Procedure in SAS (Version 9) was used in these analyses.

Because of the clustered nature of the data, axial lines were clustered within
randomly selected block groups, random intercepts and random spatial
measures effects associated with the randomly sampled block groups were
also included, to test the hypothesis that the crime counts and effects of
spatial measures on crime counts tend to randomly vary from one block
group to another.
Since space syntax measures were highly correlated with each other, it
was necessary to look at the effect of each measure on crime in a separate
model. Each model also contained sociodemographic variables. When
connectivity was entered in the model, it was positively correlated with
larceny (p<0.0001), breaking and entering (p < 0.0001), motor vehicle theft
(p < 0.0001), but not with robbery. However, additional findings showed
that the effect of connectivity on different types of crime was moderated by
390 LINDA NUBANI

levels of home ownership at the block group level. In the model, the product
of both connectivity and home ownership on larceny was negative and was
significant at p<.0001. These results suggest that the higher the percentage
of people who own their residences at a block group level, the more negative
the relationship between connectivity and larceny (see Table 1). Similarly,
models that looked at other crime types revealed that there were interactions
between the two variables connectivity and level of home ownership.
TABLE 1: Results of the model showing the effect of sociodemographic measures and
connectivity on larceny

Analysis Of GEE Parameter Estimates

Empirical Standard Error Estimates

Standard 95% Confidence

Parameter Estimate Error Limits Z Pr > |Z|

Intercept 1.4933 0.5934 0.3304 2.6563 2.52 0.0118

Connectivity 1.0976 0.2234 0.6598 1.5354 4.91 <.0001
PEROWNER -1.4959 0.6367 -2.7438 -0.2480 -2.35 0.0188
YOUTH -2.4470 3.7417 -9.7805 4.8865 -0.65 0.5131
DENSITY 0.0000 0.0001 -0.0001 0.0001 0.51 0.6099
EDUC2PER -1.5182 1.6517 -4.7554 1.7190 -0.92 0.3580
conn*PEROWNER -1.0815 0.1675 -1.4098 -0.7533 -6.46 <.0001
conn*YOUTH 0.0815 0.9203 -1.7223 1.8852 0.09 0.9295

The sample plots in Figure 2 illustrate the nature of these interactions.

Home ownership is displayed on the X-Axis and crime count is plotted along
the Y-Axis. Three regression slopes were plotted to predict larceny at
different levels of home ownership and connectivity. The unstandarized
regression coefficients were examined and were used in the model. For
example, the estimated model that was used to assess the effect of the two
independent variables (home ownership “X” and connectivity “Z”) on
larceny “Y” is as follows:

Larceny (Y) = -3.519 + 1.933 (Z) + 4.062 (X) - 2.040 (Z*X)

The methodology described here has been recommended by Aiken and West
(1991).
Perhaps these results can be related to the effects of ‘eyes on the street.’
If there are higher levels of home ownership (indicating a more stable
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 391

population), under conditions of high and moderate levels of connectivity

(supporting neighboring and ‘eyes on the street’), larceny is significantly
lower, while under conditions of low connectivity, larceny is significantly
higher. Similarly, when other crime types were examined, high levels of
home ownership in the neighborhood with high levels of connectivity
(supporting neighboring and ‘eyes on the street’), are associated with lower
levels of breaking and entering, motor vehicle theft, and robbery.
392 LINDA NUBANI

Connectivity by Home Ownership Interaction

12

10

6
Predicted Larceny

4 Low Connectivity
Count

Mean Connectivity
2
High Connectivity
0

-2

-4

-6
Low Mean High
Home Ownership

Local Integration by Home Ownership Interaction

-1

-2
Predicted Larceny Count

-3
Low Local Integration
-4
Mean Local
-5
Integration
-6 High Local Integration
-7

-8
-9

-10
Low Mean High
Home Ownership

Figure 2: Plots of interaction effects of connectivity and integration by home ownership on

larcenies

The analysis also looked at the association of global integration and local
integration with crime types and demographic measures. Global integration
was not related to any of the crime types. Local integration on the other
hand was positively linked with higher breaking and entering (p<0.0001),
higher motor vehicle theft (p=0.0139), but to not to robbery. There were
also significant interactions between local integration and home ownership
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 393

with all types of crimes. For example, motor vehicle theft, breaking and
entering, and robbery are positively associated with higher levels of home
ownership along low to moderately accessible routes. There are several
factors that may affect these results. It is perhaps less likely to be caught
along more segregated routes. Home ownership may indicate a higher level
of valuables and thus a more attractive target to criminals. Highly integrated
routes had no effects on these crimes. Larceny, on the other hand, tends to
increase with higher levels of home ownership along less integrated routes.
Fewer larcenies appeared in integrated areas with higher levels of home
ownership. A careful examination is needed to compare both sets of
interaction plots.

7. Conclusions

Results of the analysis showed that two space syntax measures, local
integration and connectivity, were highly associated with different crime
types through interactions with levels of home ownership in different block
groups. In other words, the number of intersections a route has and the
average number of turns one needs to reach it from any route in the city
objectively express the nature of its accessibility. Other factors such as
median income, youth concentration, density, racial composition and global
integration did not feature in the model. It is interesting to note that
although criminals have different motives for committing a crime whether it
is to burglarize a property or snatch a purse on the street (Davidson, 1993),
the effect of spatial measures was consistent in all types of crimes except for
larceny. Unlike other crimes types, larceny increases slightly with the
increase of levels of home ownership along highly accessible routes.
According to Brantingham (1984), if a criminal is searching for a target with
all things being equal, the closest target will be chosen. In the case of
larcenies, the effect of home ownership is weaker because larcenies are
crimes that occur to people in the streets rather in their homes. However, if
that route had more intersections, larcenies tend to dramatically drop with
higher levels of home ownership indicating higher potential movement and
eyes on the street.
To conclude, recent developments in space syntax computer software and
methodologies appear to add a promising new tool to examine the
implications of spatial layout characteristics on crime outcomes as well as
other behavioral outcomes. In this study, for example, we were able to
identify the type of streets that offered escape routes to criminals simply by
comparing the topological relationships among all the streets within a city to
actual crime counts. People usually select routes intuitively without the aid
of a map or without having an understanding of how that route is connected
394 LINDA NUBANI

to all the routes in the city. However, space syntax software enabled
researchers to understand how people behave in urban environments and
offered predictions about the streets that are more likely to be occupied by
people and the streets that are more likely to be vulnerable to crime. It is
also interesting to know that the resulting colored map may capture at a
glance the accessibility level of streets. This in turn may prove to be
valuable for police as it helps them understand where they should increase
their patrolling.
In sum, space syntax relies on advancement of computer technology to
map out spatial interrelationships and thereby allowing the researcher to
understand the structure of the city and how it is related to behavioral
outcome. It represents the physical complexity of the city as systems of
spaces created between and within buildings. This objective method is
proven powerful as it allows cities of different forms and structures to be
compared in terms of spatial interrelationships. Additionally, new urban
developments could also be tested in terms of how space will be used.

References

Aiken, L. S., & West, S.G., 1991. Multiple Regression: Testing and interpreting
interactions.
Newbury Park, CA: Sage.
Bennett, T., 1989. Burglars’ choice of targets. In D. Evans, & D. Herbert, eds. The
Geography of Crime. New York: Routledge.
Best, G. A., 1970. Direction – finding in large buildings. In: D. Canter, ed. Architectural
Psychology. London: RIBA Publications, pp. 72-91.
Davidson, N., 1993. New Directions in Environmental Criminology. In H. Jones, ed. Crime
and the Urban Environment: The Scottish Experience. Aveburg: Ashgate Publishing
Limited.
Dunn, C., 1980. Crime Area Research. In D. Georges, & K. Harris, Eds. Crime: A Spatial
Perspective. New York: Columbia Press.
Farooq, A., 1999. Social and spatial implications of community-based residential
environments on crime in urban settings (Doctoral Dissertation, Georgia Institute of
Technology). Dissertation Abstracts International, 61(04), 1196, (UMI No. 9966947).
FBI Uniform Crime Report, 1998. Retrieved October 2004 from
http://www.fbi.gov/ucr/Cius_98/98crime/98cius02.pdf.
Frank, L. D., & Engelke, P. O., 2001. The Built Environment and Human Activity Patterns:
Exploring the Impacts of Urban Form on Public Health. Journal of Planning Literature,
16 (2), 213- 218.
Gosnells Wins the Fight Against Crime, 2003. Australia’s National Government Newspaper
Online. http://www.loc-gov-focus.aus.net/editions/2003/april/gosnells.shtml
Hill, G., & Hill, K., 1995. The Real Life Dictionary of the Law. Los Angeles, CA: General
Publishing Group.
Hillier, B., & Hanson, J., 1984. The social logic of space. Cambridge: Cambridge University
Press.
Jacobs, J., 1961. The death and life of great American cities. New York: Vintage Books.
USING SPACE SYNTAX SOFTWARE IN EXPLAINING CRIME 395

Jones, M. & Fanek, M., 1997. Crime in the Urban Environment. Proceedings of Space Syntax
First International Symposium, London.
Nubani, L., 2003. The Role of Space Syntax in describing the effect of urban configuration
on pedestrian movement. Paper presented at EDRA 34, Minneapolis, Minnesota.
O’Neill, M. J., 1991. Evaluation of a conceptual model of architectural legibility.
Environment and Behavior, 23, 259-284.
Rashid, M., Zimring, C., Wineman, J., Flaningam, T., Nubani, L. and Hammash, R., 2005.
The Effects of Spatial Behaviors and Layout Attributes on Individuals’ Perception of
Psychosocial Constructs in Offices. Proceedings of the 5th International Space Syntax
Symposium, Delft, 46.1-14
Reid, S. T., 2002. Crime and Criminology. New York: McGraw Hill.
Rengert, G., 1980. Spatial Aspects of Criminal Behavior. In D. Georges, & K. Harris, Eds.
Crime: A Spatial Perspective. New York, Columbia Press.
Shu, S.& Huang, J., 2003. Spatial Configuration and Vulnerability of residential burglary: A
case study of a city in Taiwan. Proceedings of the 4th International Space Syntax
Symposium, London, 46.1-14
Siegel, L.J., 2002. Criminology: The Core. Belmont. Ca.: Wadworth Thompson Learning.
Stollard, P., 1991. Crime Prevention through Housing Design. U.K.: Spon E & F.N.
Southworth, M. & Owens, P., 1993. The evolving metropolis: Studies of community,
neighborhood, and street form at the urban edge. APA Journal, 59(3), 271-287.
Taylor, R. B., 2002. Crime prevention through environmental design: Yes, No, Maybe,
Unknowable, and All of the Above. In R. Bechtel, & A. Churchman, eds. Handbook of
Environmental Psychology. New York: John Wiley and Sons, 413-426.
Brantingham, P. and Brantingham, P., 1984. Patterns in Crime. New York: Macmillan
Publishing Company.
Ypsilanti MI Crime Statistics, 2002. AreaConnect. Retrieved July 2004 from
http://www.ypsilanti.areaconnect.com/crime1.htm