Work in our lab, the Environmental Neuroscience Lab at the University of Chicago, is beginning a new chapter in a famous theory from the early 1980s known as the broken windows theory.
Broken windows theory has been one of the most influential sociological theories of criminal behavior. The seminal ideas of the theory were proposed in an Atlantic Monthly article in 1982. James Wilson and George Kelling proposed that “disorder” in public spaces — e.g., graffiti, litter, an abandoned property, a drunk, a loiterer, a drug dealer — leads people to reason that they can get away with breaking the rules since it appears to be the descriptive norm, and perhaps such activity signals that policing and monitoring is low. Wilson and Kelling make two general predictions about the consequences of this reasoning process. First, and the more famous of these predictions, is that this sense of impunity would encourage people to behave badly — particularly those with delinquent inclinations. Second, and a more distal prediction, is that people in this community would start to feel threatened, leading to social withdraw and the deterioration of the community. In this article, when we refer to broken windows theory, we are referring to the former prediction.
One implication of the theory is that maintaining and monitoring urban environments to prevent minor rule-breaking behaviors should nip the problem in the bud, thus maintaining an environment which is orderly and lawful. The promise of this apparently simple solution to reducing urban crime motivated major reforms in criminal policy in New York City in the 80s and 90s. And indeed, this was followed by a sudden drop in both petty and serious crimes.
But post hoc ergo propter hoc. Some scholars argued that this drop was not due to “broken windows policing” but rather because of extraneous variables such as the addition and removal of lead from paint and gasoline which happened 25 years before the rise and fall in violent crime. Such research cast doubt on the causal validity of broken windows theory.
What is the best way to deal with a blow to causal validity? Conduct an experiment! In 2005, in Lowell, MA, a group of scholars from Harvard University and Suffolk University did just this. In a randomized controlled trial, they identified 34 areas high in crime. In a random half of the spots, they employed broken windows policing techniques — they cleared trash, fixed streetlights, enforced building codes, discouraged loiterers, made more misdemeanor arrests, and expanded mental health services and aid for the homeless. In the other half, no changes were made to routine police service. Broken windows policing reduced calls to the police by 20%. Interestingly, cleaning up the physical environment was more effective than increasing misdemeanor arrests and improving social services.
In 2007 and 2008, scholars at the University of Groningen published six field experiments in the journal Science that further support the causal validity of broken windows theory — that disorder spreads disorder. They found that manipulating exposure to one form of disorder (e.g., graffiti) led people to produce another form of disorder (e.g., litter). The results suggest that seeing graffiti not only encourages more graffiti, it also encourages other rule-breaking behaviors. The conclusion of this report, published in the journal Science, was that disorder spreads not just within one domain but across domains.
So there is causal evidence that broken windows may actually encourage rule-breaking, and also there is evidence that broken windows policing strategise actually have a desirable effect — particularly cleaning up the physical environment. The results are consistent with the heuristic equation put forward by the famed psychologist Kurt Lewin: Behavior = f(Person, Environment). But there is one problem. From the perspective of a cognitive scientist, the experimental evidence so far, though useful in establishing causality, is too laden with confounds to give us an idea as to how broken windows effects happen. For example, the experimental manipulations used by the above researchers manipulate not only social disorder cues (e.g., adding or removing cues to social disorder such as litter and graffiti), they also manipulate low-level visual cues (e.g., the spatial and color properties of the environment). For the human brain to make sense of social disorder cues such as graffiti and litter, it first has to make sense of basic spatial and color visual information. This earlier perceptual process is completely overlooked in research on broken windows theory. And this is where our research comes in.
In our lab we are taking a rigorous and systematic approach to studying phenomena relevant to broken windows theory. We utilize research methods from cognitive and behavioral science to investigate a sociological theory, thus bridging the gap between these disciplines.
For example, in one study, we ask, what if part of the broken windows story is not about reasoning at all? What is basic perceptual processes alone are sufficient to encourage bad behavior? To answer these questions, we use experimental and novel quantitative methods to first deconstruct and define visual disorder. We then exploited this knowledge to reconstruct basic visual disorder — absent of any social cues — and tested whether it encourages a rule-breaking behavior.
The first three experiments of this study were focused on defining visual disorder. Scholars of broken windows theory often discuss “visual disorder” without providing a concrete definition of what it is. Sometimes it sounds like they are referring to low-level visual disorder features such as spatial and color properties of an environment. Other times they seem to be referring more to high-level perception of social cues such as litter and graffiti. We wanted to define low-level visual disorder in concrete terms. In one experiment, we quantified spatial and color features of various environmental scenes and used these to predict people’s ratings of how disorderly various scenes looked. We found that spatial features — particularly the density of non-straight edges in the scene — were better at predicting the perception of disorder than color features. In two follow-up experiments, we actually extracted and scrambled the edges and colors from the scene images, thus removing the presence of recognizable objects and features (e.g., buildings, cars, trees). Non-straight edges or lines still were driving the perception of disorder. From this we could conclude that (a) real-world objects and features are not necessary for the perception of disorder and (b) at the basic visual level, spatial features matter more than color features for perceptions of disorder.
Next, we conducted three more experiments to test whether basic visual disorder can encourage rule-breaking behavior — in this case, cheating on a test. We reconstructed visual disorder by exploiting what we learned from the previous experiments. We composed images either from the most disorderly edges or the most orderly edges. We also manipulated whether the images were asymmetrical or not — a pretest confirmed that spatial asymmetry was also important for perceived disorder. To get people to cheat, we adapted a creative method to study cheating developed and validated behavioral economics expert Dan Ariely and his colleagues Nina Mazar and On Amir. In this method, people take a math test, then later grade themselves on the test. To tempt people to cheat, they are told that they will be paid a fixed amount for every question they indicate they got correct. In our version of the task, we had people view our visually disordered or visually ordered stimuli for just five minutes before grading themselves. As predicted, people who viewed the visually disordered stimuli cheated more. On average, across the three experiments using this method, the visual disorder manipulation increased the relative likelihood of cheating by 26% on average and the relativev amount of cheating by 42% on average. These two results point to two subtly different phenomena. First, the increased likelihood of cheating suggests that visual disorder encourages noncheaters to become cheaters. Second, the increased amount of cheating suggests that visual disorder encourages cheaters to cheat more. This increase in cheating behavior could have major economic and societal consequences. Imagine if the amount by which people under-reported their taxes increased by just 1% — billions of dollars would be lost.
It is becoming increasingly clear that disorderly environments encourage disorderly behaviors. But it is not clear at all why this happens, and what exactly is meant by “disorderly” environments, and for that matter, what is included in “disorderly behaviors”. The study discussed here sheds light on both of these questions. It tells us that disorderly environments may affect us because of their basic perceptual features. It also tells us that cheating is one of behaviors that may be affected by disorderly environments. Future research in our lab will shed more light on what makes an environment disorderly, what behaviors it affects, and why.
Taking a scientific approach to studying the effect of human-environment interactions on rule-breaking will shed light on the validity of the propositions of broken windows theory. It will set boundary conditions as to when the theory’s predictions are sound, and when they are not. It will also bring clarity to vague concepts such as “environmental disorder.” It will turn the page to a new chapter in which the importance of broken windows theory is met with the empirical rigor it demands.