Abstract: Although the criminal justice system is designed around the idea that individuals are invariant in their responses to punishment, research indicates that individuals exhibit a tremendous amount of variability in their punishment sensitivity. This raises the question of why; what are the individual- and situation-level variables that impact a person’s sensitivity to punishment? In the current research, we synthesize theory and research on inflammation, learning, and evolutionary biology to examine the relationship between inflammatory activity and sensitivity to punishment. These theories combine to predict that inflammatory activity – which is metabolically costly and reflects a context in which the net payoff associated with future oriented behaviors is diminished – will decrease sensitivity to punishment, but not rewards. Consistent with this hypothesis, Study 1 found that in U.S. states with a higher infectious disease burden (a proxy for average levels of inflammatory activity) exhibit harsher sentencing in their criminal justice systems. Studies 2 and 3 experimentally manipulated variables known to impact bodily inflammatory activity and measured subsequent punishment and reward sensitivity using a probabilistic selection task. Results revealed that (a) increasing inflammation (i.e., completing the study in a dirty vs. clean room) diminished punishment sensitivity (Study 2), whereby (b) administering a non-steroidal anti-inflammatory drug, suppressing inflammatory activity, enhanced it. No such changes were found for reward sensitivity. Together, these results provide evidence of a link between the activities of the immune system and punishment sensitivity, which may have implications for criminal justice outcomes.
General Discussion
In the current research, we investigated the role that the activities of the immune system play in regulating punishment sensitivity. Based on insights from research in psychoneuroimmunology (Maier and Watkins, 1998; Banks, 2005; Dantzer and Kelley, 2007; Lasselin et al., 2017; Draper et al., 2018), and RSFT (Real and Caraco, 1986; Lima and Dill, 1990; Houston, 1991; McNamara and Houston, 1992), we predicted that punishment sensitivity would decrease in contexts where inflammation is elevated and increase when inflammatory activity is diminished. This pattern was hypothesized to occur because in the context of heightened inflammation (a) an individual’s probability of survival is lower, lowering the payoffs one can expect from investing in future-oriented behaviors (e.g., Gassen et al., 2019a, b; Gassen and Hill, 2019), and (b) the immunometabolic constraints that occur in this context decrease one’s ability to inhibit dominant responses (see e.g., Lacourt et al., 2018; Treadway et al., 2019).
Preliminary support for this hypothesis was found across three studies. Study 1 revealed that an environmental factor that promotes inflammatory activity (i.e., high infectious disease burden; e.g., Zhu et al., 1999; Gattone et al., 2001; Nazmi et al., 2010; Thompson et al., 2014; Ferrucci and Fabbri, 2018) was associated with the use of harsher punishments for criminal offenses. Although there may be numerous contributors that play a role in the association between these variables, it is consistent with the hypothesis that inflammation should predict reduced sensitivity to punishment, as harsher punishments are required to modify the behavior of individuals with lower punishment sensitivity (compared to those with higher punishment sensitivity; Jean-Richard-Dit-Bressel et al., 2018; Marchant et al., 2018). In addition to providing initial support for the hypothesis that the activities of the immune system will predict meaningful differences in punishment sensitivity, these results suggest that this relationship could have implications for criminal justice outcomes.
Studies 2 and 3 found continued support for the hypothesized relationship between inflammatory activity and punishment sensitivity. Study 2 revealed that exposure to an environment that elicited increased inflammatory activity (measured via salivary IL-1β) led to diminished punishment sensitivity. The results of Study 3 found further support for this hypothesis, demonstrating that administering a manipulation designed to experimentally reduce inflammatory activity (via aspirin administration) led to an increase in punishment sensitivity. No difference in reward sensitivity was observed across these two studies. Taken together, these results suggest that the activities of the immune system – and inflammation in particular – play a role in regulating punishment sensitivity. Further, these results provide preliminary evidence that the relationship between punitive measures and infectious disease burden found in Study 1 may be driven by elevated inflammation leading to (a) decreased punishment sensitivity and (b) harsher punishments to compensate for reduced sensitivity to punishment.
Together, the results of the current research add to a growing body of work demonstrating an important role for the immune system in regulating processes involved in learning (see e.g., Depino et al., 2004; Huang and Sheng, 2010; Sartori et al., 2012). Further, the current work contributes to the body of research examining inflammatory activity and processes related to punishment sensitivity (see e.g., Pugh et al., 1998; Patil et al., 2003; Sparkman et al., 2005; Kohman et al., 2007; Harrison et al., 2016). The latter is particularly important given the inconsistent results found across previous studies using different methods. For example, some studies have found no association between states known to be associated with increased inflammatory activity and punishment sensitivity (Kunisato et al., 2012; Berghorst et al., 2013). Others have found that heightened inflammatory activity increases punishment sensitivity (Harrison et al., 2016), with participants exhibiting more punishment sensitivity on a monetary task after an inflammatory response had been elicited via typhoid vaccination. One potential explanation for the inconsistencies between this previous work and the results of the current studies is that they reflect differences in the magnitude of the inflammatory response elicited by our manipulation (i.e., Study 2; dirty room) and theirs (i.e., typhoid vaccine). The typhoid vaccination used in Harrison et al. (2016) research resulted in an average 250% increase in plasma levels of interleukin-6, a proinflammatory cytokine. Our much subtler contextual manipulation of inflammation, on the other hand, only revealed an average 191% increase (with a rather large standard deviation) in IL-1β levels in saliva. As such, one possibility is that inflammation exerts a dose-dependent effect on punishment sensitivity, where small increases in inflammation may impair punishment sensitivity (as found in the current work), and larger increases in inflammation may enhance it (as found in Harrison et al., 2016). Moreover, differences in the timing of the punishment sensitivity task between our study and Harrison et al. (2016) may also help explain the disparate findings. Participants in the current research completed the PST shortly after entering the dirty room. In contrast, the behavioral task in Harrison et al. (2016) study was administered 2.5–3.5 h after vaccination. Thus, this could indicate that the effects of proinflammatory cytokines on reward and punishment sensitivity are time-dependent. Future research should examine these possibilities.
Much of the previous research studying the impact of inflammation on punishment sensitivity has been conducted using non-human animals. Consistent with the findings reported here, this animal research suggests that inflammatory challenges often reduce performance on tasks related to punishment sensitivity, such as avoiding aversive stimuli (e.g., foot shocks or predators) and contextual fear conditioning (Pugh et al., 1998; Patil et al., 2003; Sparkman et al., 2005; Kohman et al., 2007; Adelman et al., 2017). Moreover, during acute infection, which is associated with heightened inflammatory activity, house finches exhibit reduced behavioral avoidance of predators (Adelman et al., 2017).
Inherent in the current work are several limitations. For example, although Study 1 found that states with a greater infectious disease burden exhibited harsher punishments, it is possible that this association reflects processes other than reduced punishment sensitivity in the context of heightened inflammatory activity. For example, in addition to offenders, judges in high pathogen areas are also exposed to greater infectious disease risk than those in less pathogen dense areas. Thus, high infectious disease burden may influence psychological characteristics of judges (e.g., impulsivity) that render them more oriented toward harsher sentencing. Exploring these and other possibilities will be an important direction for future research.
The experimental studies also have important limitations. While Study 2 provided evidence that our manipulation of room cleanliness resulted in heightened inflammation and decreased punishment sensitivity, results did not provide evidence that levels of IL-1β mediated the relationship between room condition and punishment sensitivity. This could be due to a variety of factors. First, we only measured one proinflammatory cytokine, IL-1β. Given that a host of different proinflammatory proteins coordinates the inflammation response, it is possible that the relationship between room condition and punishment sensitivity is driven by a proinflammatory protein that we did not measure. Second, saliva samples were collected 30 min after exposure to room condition. It is possible that participants’ levels of inflammation were declining at this time and may not have been representative of their inflammatory levels during the task. As such, this detracts from our ability to make causal inferences about the role that inflammation plays in calibrating punishment sensitivity. However, it bears noting that past research examining the influence of experimental manipulations that elicit an inflammatory response on behavior often do not test or report whether inflammation serves as a mediator (e.g., Eisenberger et al., 2010; Inagaki et al., 2012; Harrison et al., 2016). It is important that future studies report these mediation analyses to provide evidence for or against claims of causal relationships between inflammation and behavioral outcomes.
Another potential limitation of Study 2 was our measurement of IL-1β in participants’ saliva samples, as opposed to peripheral blood samples (e.g., plasma or serum). Research into the strength of correlations between salivary and plasma/serum levels of cytokines across different contexts has yielded mixed results (e.g., Cruz-Almeida et al., 2017; La Fratta et al., 2018; Lee et al., 2018), and overall, there is a paucity of research on the topic. However, our primary objective for measuring levels of IL-1β was to provide a manipulation check on the prediction that exposure to the dirty room (compared to the clean room) would lead to a rise in inflammation. Recent research suggests that salivary measures of inflammation are well-suited for this purpose (e.g., Walsh et al., 2016; Newton et al., 2017; La Fratta et al., 2018; Gassen et al., 2019a). The results of Study 2 should also be interpreted with caution given that there were unequal numbers of men and women between the two conditions. While sex was controlled for in the analyses and did not significantly interact with experimental condition to predict any outcome, this is still an important limitation to consider.
One unexpected difference in punishment sensitivity emerged between the control conditions in Studies 2 and 3. Specifically, punishment sensitivity was higher in the clean room condition of Study 2 than in the control condition of Study 3. While we cannot say for certain what accounted for these differences, there was heterogeneity in the methods and sample characteristics between the two studies that may have contributed to them. First, the testing rooms used for the control conditions in each study were not equivalent. Specifically, to increase perceptions of cleanliness in the clean room condition of Study 2, a number of steps were taken to increase the room’s cleanliness, including removing trash receptacles, wiping down all of the computers and keyboards with disinfectant wipes, and placing a large bottle of hand sanitizer near the sign-in sheet. Given that these extra steps were not taken in the second experiment (for which room cleanliness was not part of the manipulation), the room used for the control condition in Study 2 was even cleaner than that used for the control condition in Study 3. Accordingly, it is possible that differences in punishment sensitivity between the two control conditions (with higher sensitivity found in Study 2) can be attributed to greater cleanliness in the control condition for Study 2 compared to Study 3.
Further, in Study 2, before entering the experimental room, participants provided their initial saliva sample in a separate room. They were then transferred to the experimental room before completing the remainder of the study. This differs from the methodology utilized in Study 3, where the entire study was completed in a single room. Although it is unclear how these procedural differences may influence punishment sensitivity, they are worthy of note in this context. A final explanation for the differences in punishment sensitivity that emerged between these conditions could lie in differences between demographic characteristics of the samples. As is displayed in Tables 2, 3, childhood and adult SES for the sample in the clean room condition for Study 2 were higher than for the control condition in Study 3 (d = 0.37–0.40). We are not aware of extant research finding SES-based differences in performance on the probabilistic selection task, specifically. However, more generally, research finds that those from a lower SES environment exhibit a higher risk for certain behavioral problems (e.g., impulsivity: Griskevicius et al., 2011), for which reduced punishment sensitivity has been identified as part of the underlying psychological architecture (e.g., Potts et al., 2006).