III. The Development of Aggression
One of the most firmly established criminological findings is the age–crime curve, which captures the age-graded nature of delinquency. The age–crime curve resembles an inverted U, whereby delinquent involvement does not exist until around the age of 12, then rises sharply until around the age of 18 or 19, at which point it begins to decline relatively quickly. By age 30, rates of criminal involvement hover near zero and remain that way throughout the rest of life. The age–crime curve has been observed at different time periods, in different countries, and by means of different methodological techniques—it is, in short, a robust criminological finding. As a direct result, there has been little reason to suspect that the age–crime curve may not be painting an accurate picture of the ebb and flow of delinquency over time.
Part of the reason that there has been little dispute of the age–crime curve is because most criminologists study adolescents and adults but fail to investigate antisocial behaviors among children. After all, how could children commit crimes such as rape, robbery, or assault? This is, of course, a rhetorical question; children do not—in fact, they cannot— commit these types of crimes. However, they can begin to display signs of antisocial behaviors, and they can engage in various forms of aggression during the first year or two of life. The problem, however, is that this section of the life course has not been studied extensively among criminologists. In recent years, a small group of researchers, spearheaded by Richard Tremblay, have examined the use of physical aggression among children (Tremblay, 2000, 2006; Tremblay et al., 1999). Their scholarship has pointed to the possibility that theory regarding the age– crime curve may perhaps need to be revamped.
Tremblay and his colleagues have examined aggressive behavior in very young children and tracked them throughout childhood. The results of their studies have been quite striking. They have found that some children begin using aggression, including hitting and kicking, well before their first birthday; in some cases, around 7 or 8 months of age (Tremblay, 2000). Even more revealing is that Tremblay et al. have reported that more than 80% of children began using physical aggression by age 17 months (Tremblay et al., 1999). Within childhood, the peak age at which children use aggression and violence is around 2 to 3 years, after which rates of aggression decline until mid-adolescence. Other types of antisocial behaviors that are not necessarily aggressive per se are also almost universal behaviors among children. For example, Tremblay and associates (1999) found that approximately 90% of children took things from others. With age, all of these types of behaviors become less prevalent.
Against this backdrop, Tremblay and others have argued that there are really two distinct age–crime curves (Tremblay, Hartup, & Archer, 2005). The first, the traditional age–crime curve described earlier, is based on official crime measures and captures involvement in law-violating behaviors. The second age–crime curve measures not criminal involvement per se but rather physical aggression. This age–crime curve also resembles an inverted U, whereby physical aggression does not appear until around the age of 1, then increases sharply until around age 3, declining quickly thereafter. Keep in mind that this latter age–crime curve indexes only acts of physical aggression, not official acts of crime or delinquency.
The fact that there are two age–crime curves is a somewhat new finding, and the next logical question is whether these two age–crime curves are interrelated or whether they are distinct from each other. Before tackling this issue, it is first necessary to determine whether behavior is stable and what is meant by behavioral stability. There are, in general, two types of stability: (a) absolute stability and (b) relative stability. It is easiest to make the distinction between these different types of stability clear by providing an example. Suppose a group of children was examined when they were 10 months old, again when they were 18 months old, and again when they were 24 months old. Suppose, further, that at each of these three ages, they were assigned an aggressiveness score (based on a valid measure of aggression) that ranged from 0 to 10, with higher scores representing more aggressiveness.
In order for absolute stability to be preserved, the scores for each person must remain the exact same at each age. For example, if a child received a score of 3 on the aggression scale at age 10 months, that child must also receive a score of 3 on the scale at age 18 months, and he or she must also receive a score of 3 at age 24 months. Any change in the value of the aggression scale across time would reduce absolute stability. Perfect absolute stability, where everyone has the exact same score at each age, is rarely, if ever, observed in the social sciences. However, it is possible to approach perfect absolute stability in some instances. Absolute stability, in short, measures the degree to which people have identical scores on some measure over time.
Relative stability, in contrast, compares all people being assessed (on the aggression scale, in this example) and rank orders them. So, for instance, suppose that there were three children who were once again assigned an aggressiveness score (again, based on a valid measure of aggression) when they were 10 months old, 18 months old, and 24 months old. At each age, it would be possible to rank order these three children, whereby one child would be rated as the most aggressive, another would be rated as the second most aggressive, and the last would be rated as the least aggressive (or the third most aggressive). Relative stability is achieved when the rank ordering of people does not change over time. In the current example, the most aggressive child at age 10 months would also be the most aggressive child at age 18 months, and he or she also would be the most aggressive child at age 24 months. With relative stability, then, change is possible on an absolute level, as long as the rank orderings remain the same. In reality, perfect relative stability is rarely, if ever, achieved, because there is usually at least some change in the rank ordering of people.
The distinction between absolute stability and relative stability is critical, especially when one is examining behaviors over the life course. After all, the frequency with which aggression is used varies drastically across different stages of the life course, and just because someone uses aggression on a daily basis at the age of 18 months does not necessarily mean that he or she will also use aggression on a daily basis at age 35 (absolute stability); however, it is quite a different question to ask whether the most aggressive 18-month-old will mature into the most aggressive 20-year-old (relative stability). In this case, an 18-monthold might use aggression daily, which would make him or her a highly aggressive child. As an adult, however, this individual may not use aggression as frequently and instead may resort to aggression perhaps twice a month. If this is the case, then as an adult this person would still rank near the high end of the aggression spectrum. It is clear that his or her use of aggression has dropped appreciably from an absolute stability perspective, but from a relative stability perspective this person remains one of the most aggressive persons when compared with other adults. From a developmental standpoint it makes more intuitive sense to speak in terms of relative stability when one is examining the stability of behavior over time. As a result, for the most part when criminologists speak of stability they are referring to relative stability, not absolute stability, and in this research paper stability should be equated with relative stability.
A wealth of studies have examined the stability of antisocial behaviors, including violent aggression, over long swaths of the life course, and the results have been remarkably consistent. Across samples, generations, and countries, and regardless of the sample analyzed and the methodological techniques used, extremely high levels of relative stability in aggressive behaviors have been observed. In a classic article, Dan Olweus (1979) reviewed studies that had examined the stability of aggressive behavior over time; he found that aggression was extremely stable, even more so than IQ scores. Findings from more recent reviews have upheld Olweus’s original article by showing that aggressive and violent behaviors are highly stable across long periods of the life course. Persons with the highest degrees of stability, moreover, are those who score extremely high or extremely low on aggression. In other words, people who are the most aggressive (or the least aggressive) at one point in time are likely to be characterized as the most aggressive (or the least aggressive) at another point in time. Change, in other words, is highly unlikely. With this information in hand, it is probably not too surprising to learn that one of the best predictors of future criminal behavior is a history of aggressive behavior in childhood and adolescence.
Building on the studies that have examined behavioral stability, criminologists have also examined whether the age of criminal onset is associated with offending behaviors later in life (DeLisi, 2005). Age of criminal onset typically is broken down into two categories: (1) an early age of criminal onset and (2) a later age of criminal onset. Although there are many ways to operationalize an early age of criminal onset, most criminologists measure earlyonset offending by examining the age at first arrest. Offenders who were arrested at or before age 14 are usually considered early-onset offenders, whereas those who were arrested after age 14 are considered as having a later age of criminal onset.
Results culled from an impressive body of literature have shown that offenders who have an early age of criminal onset are more violent and aggressive when compared with offenders who have a later age of criminal onset. Early-onset offenders, in contrast to later-onset offenders, engage in delinquent acts more frequently; commit more serious, violent offenses; and continue to commit criminal behaviors for a longer period of time, usually well into adulthood. By all objective standards, early-onset offenders have a much worse prognosis in terms of criminal outcomes than do offenders who have a later age of criminal onset.
One of the main pitfalls of measuring early-onset offending with official crime statistics (e.g., age at first criminal arrest) is that aggressive propensities begin to emerge much earlier than the age of official offending. This is why Tremblay and others have argued that the age– crime curve does not necessarily portray an accurate picture of criminal involvement over time; instead, it ignores early childhood and pretends that this stage of the life course has no bearing on later-life aggressive conduct. With this in mind, it is now possible to return to the original question: Are the two age–crime curves related, or are they separate? According to Tremblay and others, the two age–crime curves are indeed intertwined, and in the following sections a number of different theoretical perspectives are presented that are able to shed some light on the link between early-life aggression, or the early age–crime curve, and later-life involvement in crime and delinquency, or the later age–crime curve.