The value of the dual systems model of adolescent risk-taking

In recent years, a perspective on adolescent risk-taking derived from developmental neuroscience has become increasingly popular. This perspective, referred to as the “dual systems model” (Somerville et al., 2010; Steinberg, 2010) or sometimes the “maturational imbalance theory” (Casey et al., 2011), posits that increased risk-taking during adolescence is due to a combination of heightened reward sensitivity and immature impulse control, which are tied to the development of two brain systems that undergo significant change during this age period, but that develop along different timetables. One system, which has been called the “socioemotional” incentive processing system (Steinberg, 2010; Chein et al., 2011) or “ventral affective system” (Pfeifer and Allen, 2012), is localized mainly in the ventral striatum and ventromedial prefrontal cortex. The second system, referred to as the “cognitive control” system (Steinberg, 2010; Chein et al., 2011) or “prefrontal control system” (Pfeifer and Allen, 2012), is localized mainly in lateral prefrontal, parietal, and anterior cingulate cortices (Wager and Smith, 2003; Owen et al., 2005). Briefly, the dual systems (DS) perspective posits that risk-taking during mid-adolescence is the product of the heightened reactivity of the socioemotional system against a backdrop of still maturing cognitive control. With further maturation, the socioemotional system becomes less reactive and the cognitive control system becomes stronger andmore efficient. Together, these changes lead to an increase in risk taking during adolescence followed by a decrease in risk taking as individuals move into adulthood. A recent article (Pfeifer and Allen, 2012) critiques the DS model. The authors make three main points: (1) that the data do not support the DS model because there are too few studies assessing the relationship between development in each brain system and patterns of “real-world” behavior; (2) that activation of the socioemotional system is sometimes associated with adaptive functioning, that activation of the cognitive control system is sometimes associated with maladaptive functioning, and that these pieces of evidence are contrary to the DS model; and (3) that patterns of brain development are more complex than those described by the DS theory. As proponents of the DS perspective on adolescent risk taking, we welcome the opportunity to respond to the Pfeifer and Allen critique. While, we agree that presentations of the model have sometimes oversimplified the evidence or overlooked inconsistencies in the literature, we believe that the framework continues to offer a useful model for understanding risky behavior in adolescence. In the absence of an alternative theoretical account, which Pfeifer and Allen do not offer, the DS model provides a useful heuristic for the formulation of testable hypotheses. Pfeifer and Allen (2012) make several excellent points about the model’s limitations and the challenges of mapping neuroimaging findings onto specific theoretical claims. However, in our view, there are three main shortcomings in their critique. First and foremost, the authors fail to acknowledge that there is considerable behavioral evidence consistent with the predictions of the DS model. Reward sensitivity follows an inverted Ushaped curve (Steinberg et al., 2009; Romer, 2010; Harden and Tucker-Drob, 2011). In a large behavioral study of 10to 30-year-olds, participants’ selfreport indicated a peak in sensationseeking during mid-adolescence (Steinberg et al., 2009), and on a gambling task, participants’ behavior was most influenced by rewarding stimuli during this same age period (Cauffman et al., 2010). In contrast, impulse control increases gradually and linearly, and the peak in performance on tasks measuring capabilities like planning and response inhibition occurs subsequent to the peak in reward sensitivity. This linear trajectory has been demonstrated in self-reports of impulsive behavior in several large-scale studies (Steinberg et al., 2009; Harden and Tucker-Drob, 2011). Additionally there is compelling evidence from behavioral studies of cognitive control, which demonstrate that performance improves gradually over the course of adolescence and does not peak until late adolescence (Luna, 2009; Albert and Steinberg, 2011). Furthermore, both impulsivity (Verdejo-García et al., 2008) and reward/sensation-seeking (Galvan et al., 2007; Romer, 2010) are correlated with self-reported risk-taking. Importantly, the brain systems presumed to mediate these constructs follow similar developmental trajectories. The remodeling of dopaminergic pathways connecting the ventral striatum to the PFC is most pronounced shortly after puberty, just before the rise in reward sensitivity (Spear, 2009; Luciana and Collins, 2012). In contrast, the prefrontal and parietal cortices, which are thought to support agerelated improvements in cognitive control (Luna and Sweeney, 2001; Bunge et al., 2002; Astle and Scerif, 2009; Luna et al., 2010), are among the last brain regions to mature (Huttenlocher, 1990; Giedd et al.,