Examining the Direct and Indirect Effects of Fear and Anger on Criminal Decision Making Among Known Offenders

Abstract

Deterrence represents the central theoretical core of the American criminal justice system, yet relatively little attention has been paid to how emotions like fear and anger may relate to deterrence. Psychological research has debated whether negative emotions each have similar impacts on decision making (valence approaches) or if distinct emotions have unique impacts (appraisal tendency approaches). This study explores the direct and indirect influences of fear and anger on hypothetical drunk driving likelihood, including their impact on cost perceptions. Surveys were administered to 1,013 male and female incarcerated felony offenders in the Southwestern United States. Using a multivariate path model and controlling for a number of other individual factors, current fear related to increased cost perceptions and anger to decreased costs. Anger also maintained a direct influence on drunk driving, whereas fear did not. Despite their shared negative valence, fear and anger appear to have dissimilar influences on cost perceptions and criminal decision making. A better understanding of these processes may lead to improved crime prevention approaches.

Keywords

fear anger emotion decision making deterrence

Introduction

Deterrence theory is the central theoretical premise on which the criminal justice system in the United States operates (Nagin, 2007; Pratt & Cullen, 2005; Stafford & Warr, 1993), and yet relatively little attention has been paid to the issues of choice and free will in criminology (Cullen, 2011; Nagin, 2007). Nagin (2007, p. 259) called for “moving choice to center stage in criminological research and theory” and also for increased attention to how emotional/visceral states may impact those choices. Cullen (2011) reiterated this call for attention to criminal decision making, and in particular the role of emotions in the process (see also Karstedt, Loader, & Strang, 2011, for other recent calls for attention to the role of emotion in crime, and criminal justice). While a number of at least modestly optimistic reviews of the deterrence literature exist (Blumstein, Cohen, & Nagin, 1978; Nagin, 1998; Paternoster, 1987), others have recently questioned the utility of deterrence theory as an explanation of crime (Pratt & Cullen, 2005).

A more complete understanding of how emotional states like fear and anger may influence offending decisions (as outlined by deterrence or rational choice theory) may be one means of increasing the effectiveness of efforts to prevent or respond to crime. For instance, Cusson (1993) suggested that perception of situational risks can cause offenders to refrain from their intended crimes, because at the critical moment they are overcome by feelings of fear. It is possible that these kinds of effects could be exploited in efforts to increase the effectiveness of deterrence policies and intervention. Clarke and his colleagues (Clarke & Felson, 1993; Cornish & Clarke, 1987) have suggested that accurate understanding of criminal decision making must include consideration of the role of individual and situational factors, including emotions, within the decision. In addition, Clarke and Homel (1996) have applied this perspective to the development of situational crime prevention strategy that would increase the experience of negative emotions (i.e., shame and guilt) to increase the effectiveness of situational deterrence. While fear may serve to deter offending, at least partly by altering cost perceptions, the experience of anger has been shown to reduce the influence of costs on decisions (Baron, 1973, 1974, 1979; Carmichael & Piquero, 2004; Zillman, 1979) and thus increase the likelihood of aggressive behaviors through a number of psychophysiological mechanisms (Novaco, 2011).

In psychology, research on emotions’ impact on decision making has taken two distinct approaches. Initially, researchers considered the valence of the emotion (either negative or positive) and proposed that many negative emotions (e.g., fear and anger) would lead to more cautious decision making, and conceivably lower crime likelihood (see Forgas, 1989, for an example; see also Loewenstein & Lerner, 2003, for a review of this early perspective). The other, more recent approach, suggests that specific emotions will have a unique influence on decision making, sometimes referred to as an “appraisal tendency” (see Lerner & Keltner, 2001, for a thorough discussion). The current study investigates whether two negative emotional states (fear and anger) have similar or distinct direct and indirect (by altering cost perceptions) effects on the hypothetical likelihood of drunk driving, using a sample of over 1,000 convicted felony offenders.

Emotional Influences on Decision Making

Bouffard, Exum, and Paternoster (2000) proposed that emotional (e.g., anger, fear) and visceral states (e.g., hunger, sexual arousal) could have important impacts within the decision-making process. Similarly, Peters, Vastfjall, Garling, and Slovic (2006, p. 81) refer to “emotions as a spotlight,” which focuses one’s attention on certain factors within the decision. Loewenstein (1996) outlined the ways in which immediately experienced emotional states (and visceral states, like hunger) can serve to focus an individual’s attention on their own needs, on the present time period, and on factors that are consistent with the particular state (e.g., the prospect of food when hungry). These types of emotional/visceral states also serve to limit the attention paid to other considerations (e.g., concern for others). Specifically, Loewenstein and Lerner (2003, p. 269) state that emotions can cause decision makers to “selectively attend to, encode and retrieve emotion-relevant information,” which then influences the ultimate decision.

According to Loewenstein and Lerner (2003), some early research on the impact of emotional valence (i.e., negative vs. positive emotions, like anger and happiness, respectively) suggested that negative emotions in general predisposed more cautious deliberation (e.g., Forgas, 1989; Johnson & Tversky, 1983, as cited by Loewenstein & Lerner, 2003). Tiedens and Linton (2001) summarized early research on the experience of various moods and emotions as suggesting that “people engage in more systematic processing when in negative emotional states or moods, where people in positive moods or emotional states engage in more heuristic processing” (p. 976).¹ In a physiological sense, Loewenstein and Lerner (2003) suggest that these kinds of negative emotions in general might serve to alert the body that something in the environment needs attention, as opposed to positive emotions that would indicate a sense of well-being. From the emotional valence perspective, negative emotions in general would lead one to engage in more cautious decision making and thus perceive higher costs whether the individual is either fearful or angered.

More recently however, researchers (e.g., Lerner & Keltner, 2000) have suggested that specific emotions have unique effects on decision making independent of their valence. One emotional dimension that has been speculated to influence decision making is certainty (see Frijda, 1969, and Smith & Ellsworth, 1985, for a discussion of other dimensions on which emotions can vary). For example, Tiedens and Linton (2001) demonstrated that emotions characterized by certainty (i.e., the individual understands what is happening and can predict what will happen) including anger and happiness, lead to quick, heuristic decision making, whereas “uncertain” emotions (e.g., fear, sadness, worry) evoked more careful deliberation.

This consistency between emotional state and decision processes is described as an “appraisal tendency” (Lerner & Keltner, 2000, p. 476). In support of the “appraisal tendency” view, Lerner and Keltner (2001) found that when evaluating possible courses of action, fearful individuals gave more negative assessments of risks, and made more risk-averse choices. Conversely, individuals experiencing another negative emotion— anger—exhibited more risk seeking. In regard to the appraisal tendency model, Lerner and Keltner (2001) suggest that researchers should investigate whether different emotional states have direct or indirect effects on decision, and examine how information processing may differ with current emotional state. Specifically, whether angered individuals perceive lower risks and costs, and currently fearful individuals have higher perceptions of costs.

Loewenstein and Lerner (2003) also outline how the intensity of the emotional state influences the manner in which the emotion impacts the decision. For instance, at low levels of intensity, emotions can serve a simple “advisory” function, whereby they provide the individual with a kind of “gut reaction” to potential course of actions, which can serve to streamline some kind of decisions (see also Damasio, 1994). At low to moderate levels of emotional intensity, the decision maker may then focus more attention on factors that will quickly resolve the state (i.e., higher benefit estimates and lower cost perceptions). Under the influence of high-intensity visceral states, the individual’s cognitive deliberation may actually be overwhelmed, leading to behavior that goes against the individual’s own self-interest. Extreme levels of fear may, for instance, overwhelm the person to the point they are “paralyzed by fear,” incapable of engaging in any sort of self-protective behavior at all. Likewise, at extreme levels of anger, the individual may not “give a damn about” the potential consequences of offending, and thus act aggressively (Zillman, 1979, p. 279).

Considerable research on the role of emotions and visceral factors in decision making has been conducted in other fields (e.g., human decision making, psychology, economics), with criminological theorizing only recently beginning to recognize the role of emotions in offender decision making (Cullen, 2011; Nagin, 2007). A number of criminological studies have integrated emotions into perceptual deterrence and rational choice theories, especially the visceral state of sexual arousal (Ariely & Loewenstein, 2006; Bouffard, 2002a, 2011; Loewenstein, Nagin, & Paternoster, 1997) as it influences the perception of various crime benefits (e.g., more anticipated sexual pleasure among those experiencing states of sexual arousal). To date, however, very little, if any, research has examined the influence of current fear on the perceived costs of crime. Likewise, only a few studies have examined the role of anger in cost perceptions (Carmichael & Piquero, 2004; Exum, 2002).

Criminological examinations of the role of emotion and visceral factors in decision making have tended to experimentally manipulate the individual’s current state (Ariely & Loewenstein, 2006; Bouffard, 2002a, 2011; Exum, 2002; Loewenstein et al., 1997). One study (Carmichael & Piquero, 2004) did rely on the individual’s imagined level of anger if they were in a situation like the one described in the hypothetical scenario. These authors found that formal and informal cost perceptions were not deterrent among those reporting very high levels of imagined anger, whereas these costs did deter participants who imagined being less angry in this kind of situation. Likewise, Exum (2002) found that a combination of experimentally induced alcohol intoxication and anger led to higher estimates assault likelihood, but this effect was not mediated by cost perceptions, instead appearing to be a direct effect on offending.

Current Focus

Based on prior research on the “appraisal tendency” model (Lerner & Keltner, 2001), the current study examines whether fear and anger influence hypothetical drunk driving, either directly or indirectly, by altering cost perceptions. In other words, do these two negative emotional states impact the decision-making process by altering cost perceptions? Specifically, the first research question asks whether these two emotions have indirect effects on hypothetical drunk driving, by altering individuals’ cost perceptions. A number of recent studies in criminology have examined how sexual arousal, anger, and intoxication may influence the decision-making process (using mostly student samples), but how anger, and especially fear, may influence this process has not been thoroughly examined. In addition, based on past research on appraisal tendencies, it is expected that fear would lead to increased cost perceptions (more deliberative decisions), whereas anger may either indirectly influence offending by decreasing cost perceptions or have a direct influence on offending (bypassing cost considerations altogether). Thus, the second research question asks whether these direct and indirect effects are similar for states of fear and anger.

Unlike many of the studies of emotion and criminal decision making in criminology, the current study also has the added benefit of examining a large sample of serious offenders (i.e., recently incarcerated felons), many of whom (about 84%) have experience with actual drunk driving. The current study makes use of naturally occurring variation in the surveyed inmates’ current emotional states, rather than experimentally manipulating them. Although this approach has yet to be used in criminology, it has been successfully utilized in other fields. For instance, Lerner, Gonzalez, Small, and Fischhoff (2003) found that both experimentally manipulated and naturally occurring levels of fear were related to estimates of risk from terrorism among a nationally representative sample of over 900 Americans.

Methods

Procedure

Data were collected from a convenience sample of 1,013 male and female convicted, felony-level offenders. These offenders had recently been admitted to one of two intake facilities (one for men and one for women) in a large Southwestern state, and were invited to participate in an institutional review board–approved study described simply as one of “decision making.” Inmates were asked to participate in the survey during an orientation class completed by all individuals during their first week in their respective intake prisons. Inmates typically stay in these facilities for as long as two weeks, for assessment and classification prior to being transferred to the correctional facility where they will eventually serve out their sentence. These orientation classes occur daily at the men’s facility and one to two times per week in the women’s prison.

Between the months of January and May, 2011, two female and one male research assistants (working in teams of two or three) administered the surveys to inmates in 35 male orientation classes, and in nine similar classes in the women’s intake facility. Classes surveyed were not randomly selected; rather, they were selected based on the availability of these research assistants to conduct surveys. Approximately two classes per week were surveyed at the men’s facility and classes at the women’s facility were surveyed approximately once every 2 weeks (due to the longer distance between the university and the women’s prison). The researchers read the survey out loud to assist any inmates who may have had difficulty reading.

Prison system policy prohibited the compensation of inmates for their participation in the research. Despite this, a relatively high response rate of 83% was obtained (i.e., 1,013 of 1,223 offenders solicited). Offenders were informed that neither their decision to participate, nor any of the information they provided would affect their standing at the correctional facility and that the surveys would be completed anonymously. A Spanish language version of the survey was available and those surveys completed in Spanish (n = 37) were translated into English after data were collected (by a fourth, Spanish-speaking graduate student from Mexico).

Participants

Offenders were selected from two separate facilities such that the demographic information presented in Table 1 is presented separately for males and females, to assess how well each sample represents its corresponding prison’s population. Analyses presented later include both the combined male and female offender samples, as well as some brief, supplemental analysis of each gender sub-sample. In general (and despite the use of a convenience sampling strategy), each sample represented its respective prison population fairly well in terms of age, current offense, and race. The one exception is the apparent under-sampling of White inmates in the male facility (about 42% in the sample, 59% in the men’s prison). The average age of the combined male and female sample was 32 years (see Table 1) and overall about 47% described themselves as White. In addition, the sample is also characterized by a wide range of current felony offenses ranging from drug and driving while intoxicated (DWI) offenses to property offenses (e.g., burglary) and crimes of violence, including robbery, aggravated assault, and homicide. As can be seen in Table 1, these offenses reflect the range of offenses seen in each prison population relatively well. Approximately 72% of the overall sample reported that they had previously driven drunk without being caught, and the combined sample had an average of seven prior adult arrests. In terms of their self-reported likelihood of drunk driving, the average likelihood was about 45%, although 75.2% of offenders reported some (i.e., a non-zero) probability of driving drunk.

Table 1.

Characteristics of the Sample.

	All offenders	Male offenders		Female offenders
	Sample	Sample	Prison	Sample	Prison
Age (SD)	32.2 (9.9)	32.0 (9.9)	34.6	33.4 (10.1)	35.7
% White	46.9	41.8	59.0	66.5**	64.7
Current offense type
% Drug/DUI	38.4	38.9	29.7	36.2	37.6
% Violent offense	32.4	35.3	43.7	19.1	21.1
% Property offense	15.3	14.2	16.8	20.4	29.8
No. of adult arrests (SD)	7.2 (8.0)	7.52** (8.4)	—	5.98 (5.8)	—
% Ever driven drunk	83.5	85.6**	—	73.5	—
% Driven drunk and not been caught	71.6	72.7		66.5
Self-reported likelihood of driving home drunk in scenario (SD)	45.3 (36.3)	47.0** (35.9)	—	37.9 (37.0)	—
Self-control (SD)	2.27 (0.73)	2.3 (0.72)	—	2.2 (0.80)	—
Average fear score (SD)	1.95 (0.92)	1.85 (0.84)		2.43** (1.12)
Average anger score (SD)	1.82 (0.90)	1.81 (0.92)		1.87 (0.84)

Significantly higher among this gender, p < .01.

Measures

Drunk driving likelihood

The outcome variable utilized in this study is the participant’s self-reported likelihood of driving home drunk solicited in response to a hypothetical scenario. A drunk driving scenario was chosen because this offense type does not require any special knowledge, experience, or expertise. As such, many types of offenders should be able to realistically imagine engaging in this crime. This appears to have been true in that nearly 84% of the overall sample had engaged in this behavior previously. At the same time, drunk driving does represent a relatively serious offense, compared with the minor forms of crime or deviance (i.e., digital piracy or cheating) that have been examined in some studies that use student samples (e.g., Higgins, Wilson, & Fell, 2005; Tibbetts & Myers, 1999). Given that little, if any, existing research has examined these issues, in particular among a sample of known offenders, results related to even a single crime type have the potential to advance our knowledge in this area. The hypothetical scenario is also similar to those used in numerous past studies of deterrence, rational choice, and the role of emotions (Bouffard, 2002b; Loewenstein et al., 1997; Nagin & Paternoster, 1993; Nagin & Pogarsky, 2001; Piquero & Bouffard, 2007; Piquero & Tibbetts, 1996; Pogarsky, 2002) and thus adds to the existing knowledge in this specific area.

The scenario (see the Appendix) describes an individual who needs to get home after drinking at a party, but also recognizes that he or she needs the car for an appointment in the morning. Participants were instructed to read the story as if they were in the scenario and to answer a series of additional questions related to it. Participants were asked to rate their likelihood of driving home (using a 0% = not at all likely to 100% = very likely scale) even though they suspect they are over the legal limit for driving under the influence.² Participants were also asked how clearly they could imagine themselves in the situation described in the story, as well as how realistic they found the scenario to be. Participants responded to these clarity and realism questions using a similar 0% = not at all to 100% = very likely scale. The average realism rating for the overall sample was 92% and the average clarity rating was 62%.

Perceived consequences

After reporting their hypothetical likelihood of driving home, participants were instructed to develop a list of up to seven negative consequences that they thought might occur if they drove drunk as described in the scenario. Specifically, they were asked to self-generate a list of consequences using the method developed by Bouffard (2002b).³ The sample reported an average of 3.42 (SD = 1.7) total cost items. Only 43 (4.2%) inmates reported no costs for drunk driving and these cases were omitted from the later analyses, because they had no ratings for cost certainty and so on (i.e., it is not possible to model the relation between fear and cost certainty if no costs/certainty scores were reported).⁴

After listing the costs they perceived as possibly resulting from this behavior, participants were asked to indicate how likely they thought each of the consequences was to happen (perceived certainty). Participants were also asked to rate “how bad would it be” if each listed consequence did happen (perceived severity). Finally, participants were asked about the perceived salience of each cost (i.e., “How important is each item to your decision?”). Perceived certainty, severity, and salience ratings all used a similar (0% = not at all to 100% = very) scale.

Following the procedure utilized by Bouffard (2002b), these cost responses were first coded into appropriate cost type categories (e.g., “get arrested” was categorized as a “legal” cost). If an individual reported more than one item for a given cost type (e.g., two legal items, “get arrested” on one blank line, and “go to jail” on another), the average of these items’ certainty scores was calculated to represent that participant’s “legal certainty” variable, and similarly the average of these (legal) items’ severity and salience scores were used to represent that person’s “perceived legal severity” and salience scores, respectively. In addition, the average certainty of “all reported costs” was calculated as the average of all certainty scores for all cost items reported by each individual participant (the same process was used to calculate the average severity and salience for all reported costs, as well).

Once categorized, it was determined that the most often reported cost type was accident costs (77.9%), consisting of items such as “wrecking the car” or “damage other’s property.” Legal costs were reported almost as often (73.2%), consisting of items such as “get pulled over” or “get arrested.” If a response indicated that driving drunk would injure or kill another person, it was coded as “injure others” (53.2%). In addition, 41.2% of the offenders (n = 417) reported that they might kill or injure themselves if they drove drunk (these were responses that were distinct from the “accident” and “injure others” categories, such that there is no overlap in the content of items in these three groups). Finally, a number of other cost types were reported by considerably smaller numbers of participants. For instance, only 2.9% of the offenders (n = 29) reported potential family consequences (e.g., “family is disappointed”). Similarly, only about 1.3% of participants (n = 13) reported responses indicating costs to peer relationships (e.g., “reputation is ruined”). Bivariate statistics presented later include only the four most commonly reported cost types, legal, accident, injure self, and injure others, because of the small samples for other reported costs.

To make use of the entire sample of offenders (rather only those individuals who reported a particular cost type), multivariate path models will include only aggregated, all cost measures for certainty, severity, and salience. This process of aggregating numerous, diverse cost types into a single cost measure has been used in numerous past studies (see Bachman, Paternoster, & Ward, 1992; Carmichael & Piquero, 2004; Loewenstein et al., 1997, for examples). This process is also consistent with the rational choice formulation originally offered by Becker (1968) whose mathematical formula for expected utility includes the set of punishments the individual considers. Cronbach’s alphas range from .7 for all cost severity to .9 for all cost salience, suggesting good internal consistency for these aggregated cost scales.

Self-control

After completing the hypothetical drunk driving questions, participants were asked to complete the 24-item Self-Control Scale (Grasmick, Tittle, Bursik, & Arneklev, 1993), which included items such as “I often act on the spur of the moment without stopping to think” and “I like to test myself every now and then by doing something a little risky.” These items were rated using a 5-point Likert scale (4 = strongly agree, 2 = no opinion, 0 = strongly disagree). Item responses were then reverse coded to ease interpretation, such that higher scores now indicate higher levels of self-control. The average score on these 24 items was used to represent the individual’s level of self-control (Cronbach’s α = .884). Among all offenders, the average self-control scale was 2.27, with no differences by sex (see Table 1).

Other covariates

Participants provided information on their demographic characteristics (age, sex, race/ethnicity) and they were also asked to report whether they had previously engaged in drunk driving, and whether they had done so without being caught (each scored 0 = no, 1 = yes). The drunk-driving-without-being-caught variable specifically was used to control for the “experiential effect” described by Paternoster, Saltzman, Waldo, and Chiricos (1983), in which individuals’ experiences with offending and getting away with it have been demonstrated to negatively influence their later perceptions of the certainty of costs occurring in the future (see also Stafford & Warr, 1993, for a discussion of the impact of having avoided punishment on future offending).

Positive and Negative Affect Schedule (PANAS)

Participants also completed the PANAS, a widely used, 20-item scale to assess the individual’s current affective state (Watson, Clark, & Tellegen, 1988). Ten of the items on the PANAS are designed to capture positive-valence affective states (e.g., excited, proud), with the others intended to capture negative affective states (e.g., afraid, hostile). In their original study of the schedule, Watson and colleagues (1988) determined that the positive and negative affect scales were uncorrelated with one another, exhibited good internal reliability, and were stable over a 2-month follow-up period. The time frame within which individuals can be asked to respond to each item can be varied, from “How do you feel, at this moment” up to “Have you felt this way during the past year” (Watson et al., 1988, p. 1070). Individuals in the current sample were asked to report how they felt “right now, that is, at this moment,” such that their answers reflect their current mood state at the time of the survey, rather than some trait emotion (see Tibbetts, 2013, for a discussion of state vs. trait emotions). Participants responded to each item using a 5-point scale, with 1 labeled very slightly or not at all and 5 labeled extremely.

A factor analysis (principal components analysis, with varimax rotation) on the current sample of offenders revealed the presence of five distinct factors with eigenvalues over 1.0 (several other rotational strategies, such as direct oblimin, which allow for correlated factors, were also attempted and yielded the same set of five factors). Rotated factor loadings (from the varimax rotated factor solution) for each item are presented for each of the five factors. Factor 1 (labeled “fear,” eigenvalue = 4.575, explained variance = 22.88%) included the items “distressed” (.554), “scared” (.830), “nervous” (.819), “jittery” (.592), and “afraid” (.840). The second factor (“determined,” eigenvalue = 3.599, explained variance = 18.00) included the items “strong” (.588), “alert” (.728), “inspired” (.570), “determined” (.813), “attentive” (.827), and “active” (.605). Factor 3 (“excited,” eigenvalue = 1.819, explained variance = 9.10) contained the items “interested (.583), “excited” (.846), “enthusiastic” (.839), and “proud” (.610). The fourth factor (“anger,” eigenvalue = 1.596, explained variance = 7.98) contained the items “upset” (.563), “hostile” (.818), and “irritable” (.836). Finally, Factor 5 ("shame", eigenvalue = 1.032, explained variance = 5.158) included the items “guilty” (.803) and “ashamed” (.743). Based on previous research examining these two negative emotions, scale scores for fear and anger are of particular interest in the current effort, and the measures of these two states used in subsequent analyses are the simple average of each set of specific items (e.g., the average of the five fear items for the fear scale). In addition, each of these two scales exhibited acceptable levels of internal reliability; Cronbach’s alpha for the five-item fear scale was .82, and for the three-item anger scale it was .66.

As the PANAS was administered after the participants had considered their hypothetical likelihood of driving drunk and the consequences they perceived as possibly resulting from such behavior, it is possible that the individual’s current emotional state may have been altered by these questions, which had preceded the assessment of emotional state. Several factors suggest that this was not the case. First, it is difficult to imagine that convicted felony offenders, who have considerable experience with actual offending and who have recently been admitted to a large state prison (i.e., stressful environment), are made appreciably more fearful after having been asked to consider a hypothetical drunk driving scenario.

Several statistical findings also support this supposition. For instance, if considering the consequences of hypothetical drunk driving increased levels of fear in participants, one would expect fear to be higher among those who more clearly imagined themselves in the hypothetical scenario, or those who found the scenario to be particularly realistic. In fact, neither the correlation between clarity and fear (r = .03, ns) nor realism and fear level (r = .04, ns) was significant. Likewise, individuals who consider the consequences of drunk driving, and then become fearful as a result, would probably also report higher numbers of costs, although again there is no significant correlation between the number of costs reported and level of fear (r = .01, ns). Finally, if considering the hypothetical decision induces fearfulness, those who experience high levels of fear when asked to make this consideration should also have reported lower drunk driving intentions, but the correlation between fear and drunk driving is not significant (r = −.024, ns).

On the other hand, drunk driving likelihood is significantly correlated with scores on the anger scale (r = .159, p < .01). Likewise, those who reported some (greater than 0%) probability of hypothetical drunk driving reported higher levels of current anger (mean anger score = 1.87) than did those who said they would not drive drunk (mean anger score = 1.63, t = −3.310, p < .01). It is difficult to conceive of an explanation for why deciding that one has a high drunk driving likelihood would then lead that individual to experience heightened feelings of anger (i.e., being upset, hostile, and irritable). Overall then, there is little empirical evidence to support the counterargument that considering possible costs and hypothetical likelihood of drunk driving influence current emotional state, rather than the other way around.

Planned Analyses

First, bivariate analyses will briefly examine the relationships between each emotional state (fear, anger) and perceived costs, and offending likelihood. Then, path models will be used to examine whether the current emotional states of fear and anger are indirectly related to offending likelihood through altered aggregated cost perceptions, controlling for a number of individual factors. Results presented in the previous section suggest little reason to suspect that the hypothetical offense decision itself impacted current emotional state (i.e., a reciprocal effect of cost perceptions on current emotional state), so reciprocal paths are not included.

Results

Bivariate Analyses

Results presented in Table 2 demonstrate that being fearful is significantly positively correlated with the average certainty of all reported costs (r = .150, p < .01). Scores on the fear scale are also significantly correlated with the certainty of legal costs (r = .161, p < .01), accident-related costs (r = .117, p < .01), and injure-self type costs (r = .168, p < .01), whereas fear is not significantly related to the perceived certainty of injuring others (r = .072, ns). Fear scores are also significantly related to the higher perceived salience of all reported costs (r = .088, p < .01), legal costs (r = .107, p < .01), and accident-related costs (r = .120, p < .01), but not to the salience of costs related to injuring oneself or others. None of the severity scores for the specific cost types examined here exhibit significant correlations with the level of fearfulness, but the average perceived severity of all reported costs does have a small but significant positive relationship with fearfulness (r = .080, p < .05). All cost certainty was also correlated with all cost severity (r = .243, p < .01) and salience (r = .166, p < .01), while severity and salience were also significantly correlated (r = .276, p < .01).

Table 2.

Correlations Among Emotion States and Variables of Interest: All Cases.

	Fear	Anger
All cost certainty	.150**	.065^†
All cost severity	.080*	−.043
All cost salience	.088*	−.079*
Legal certainty	.161**	.049
Legal severity	.070	−.025
Legal salience	.107**	−.029
Accident certainty	.117**	.024
Accident severity	.058	−.032
Accident salience	.120**	−.078^†
Injure self certainty	.168**	.061
Injure self severity	.005	−.098
Injure self salience	−.056	−.039
Injure others certainty	.072	.072
Injure others severity	.050	−.043
Injure others salience	.034	−.057
Self-control	−.065^†	−.309**
Risk seeking	−.035	—
Temper	—	.367
Drunk driving likelihood	−.024	.159**
Age	.097**	−.132**
Anger	.407**	—

†

p < .10. *p < .05. **p < .01.

Many of the bivariate relationships between the fear scale and cost perceptions are relatively small in magnitude, but reach statistical significance (likely due at least in part to the large sample size). Unlike other studies of the role of emotion in criminal decision making, recall that this study takes advantage of naturally occurring variation in current affective states (rather than an experimental manipulation of the participant’s emotional state), which may partially explain the smaller size of these relationships. The level of current fearfulness did not differ between those who reported some (greater than 0%) hypothetical likelihood of drunk driving (fear score = 1.93) than did those who reported a zero probability of driving drunk (fear score = 1.98, t = .645, ns).

Anger on the other hand does not have a significant correlation with either aggregated costs’ certainty or severity scores (see Table 2), nor with any of the certainty or severity scores for specific cost types. Anger is related to significantly lower perceived salience of all reported costs (r = −.079, p < .05), but none of the other two aggregated cost measures. Negative correlations between anger and the salience of the three other specific cost types examined here (legal, injure self, injure others) also did not reach statistical significance.

As would be expected from their shared negative valence, anger and fear were significantly, positively correlated with each other (r = .407, p < .01). Despite this shared negative valence, these correlational results seem to suggest that these emotions have different relationships with cost perceptions. In addition, levels of fear appear unrelated to the likelihood of driving home (r = −.024, ns). Conversely, angered individuals were significantly more likely to say that they would drive drunk (r = .159, p < .01), suggesting that these two states may also have different direct relationships with offending likelihood. Older individuals reported significantly higher levels of fear (r = .097, p < .01), and lower levels of anger (r = −.132, p < .01).

Finally, scores on the Grasmick et al. (1993) Self-Control Scale were significantly negatively correlated with anger (r = −.309, p < .01), but not with fear (r = −.065, ns). These correlational results suggest that these measures of emotional state (especially current anger) may be partly (but not primarily) related to longstanding attitudes or tendencies to experience these states (i.e., temper). Recall for instance that the Grasmick et al. (1993, p. 14) scale includes four items representing one’s temper (e.g., “I lose my temper pretty easily”) and four other items representing one’s preference for risk seeking (e.g., “Sometimes I will take a risk just for the fun of it”). In fact, the correlation between the average of these four temper items and the PANAS anger scale is significant (r = .367, p < .01), suggesting that about 13% of the variance in PANAS anger scores is accounted for by one’s temper. On the other hand, individuals who generally prefer risk seeking would be expected to have lower levels of fear in a given situation. In fact, the correlation between the PANAS fear scale and the average score on the four risk seeking items is not significant (r = −.035, ns). Overall, while some portion of one’s current state of anger (about 13%) may be explained by temper, very little of the individual’s state of current fearfulness appears to be explained by his or her tendency to prefer (or be averse to) risky behaviors.

Multivariate Path Models for Fear

A multivariate path model was estimated, using AMOS 20.0 to examine the possible direct and indirect effects of fear and anger on drunk driving likelihood. This model included direct paths from fear and anger to drunk driving likelihood, as well as indirect paths from each emotional state through each cost perception measure (all cost certainty, severity, and salience). The initial model also included similar direct and indirect effects (through cost perceptions) of age, sex, minority group status, overall self-control scale scores, and the experiential effect (see Figure 1 for this initial model containing all originally included direct and indirect pathways). Finally, because anger and fear were significantly correlated with one another, their error terms were allowed to covary in this path model, as was the case for the three cost perception measures, certainty, severity, and salience. This initial model with all possible direct and indirect pathways specified failed to provide a good fit to the underlying data. For instance, the ratio of χ² to degrees of freedom = 10.878, Critical N (CN) = 196, comparative fit index (CFI) = .932, root mean square error of approximation (RMSEA) = .099; generally, ratio of χ² to degrees of freedom should be less than 5, CN should be greater than 200, CFI should be close to 1.0, and RMSEA should be less than .06, but may be as high as .08 to indicate a reasonable fit; see Arbuckle, 2011).⁵

Figure 1.

Initial model of influences of fear and anger on aggregated cost perceptions and drunk driving.

Figure 2 depicts the final model that was fitted to the underlying data, after the removal of various non-significant paths. In general, the model provided goodness-of-fit statistics that were considerably improved over Model 1 (ratio of χ² to degrees of freedom = 4.518, CN = 346, CFI = .911, RMSEA = .059). A model comparison test revealed that the model presented in Figure 2 provided a significantly better fit to the data; the difference between the chi-square values from the two models was 33.118 and the difference in the number of degrees of freedom was 16 (p < .01). These results generally support the “appraisal tendency” perspective in that fear was related to higher cost perceptions and anger was related to decreased cost perceptions, whereas higher cost perceptions predicted lower intention to drive drunk. Specifically, level of fear was significantly, positively related to the average perceived severity (Standardized Estimate [Std. Est.] = .124, p < .01) and salience (Std. Est. = .147, p < .01) for all reported costs, but not to cost certainty. Anger scores however predicted significantly lower perceptions of all cost severity (Std. Est. = −.108, p < .01) and salience (Std. Est. = −.141, p < .01). Overall, as anticipated, it appears that fear increased the severity and salience of cost perceptions, whereas anger had the opposite effect in that it reduced the individual’s perception of costs as severe or salient. In terms of factors influencing the hypothetical decision to drive drunk, perceived salience of all costs was significantly negatively related to the likelihood of driving drunk (Std. Est. = −.099, p < .01), as was the perceived certainty of all costs (Std. Est. = −.141, p < .01). Finally, anger maintained a significant direct effect on the hypothetical likelihood of drunk driving (Std. Est. = .083, p < .05), whereas fear did not directly impact drunk driving intentions.

Figure 2.

Final model of influences of fear and anger on aggregated cost perceptions and drunk driving: All cases.

Beyond these theoretically expected direct and indirect effects of emotions on drunk driving intentions, a number of other significant influences on cost perceptions were also observed. For example, minority group members (Std. Est. = .124, p < .01) reported higher cost certainty scores, whereas males (Std. Est. = −.175, p < .01) and those who had driven drunk without being apprehended (Std. Est. = −.104, p < .01) reported lower certainty scores. Finally, a number of direct effects on hypothetical drunk driving also emerged. For instance, higher self-control scores were related to significantly lower drunk driving likelihood (Std. Est. = −.240, p < .01). In addition, those individuals who had driven drunk in the past without being apprehended (Std. Est. = .167, p < .01) and male inmates (Std. Est. = .069, p < .01) both reported significantly higher likelihood of driving home.

Gender-specific supplemental analyses

The final model presented in Figure 2 was replicated on the separate sub-samples of male and female inmates to investigate any possible gender differences in the operation of emotional states on decision making. The male-only model produced reasonable goodness-of-fit statistics (ratio of χ² to degrees of freedom = 3.425, CN = 384, CFI = .931, RMSEA = .054). This model estimated on the male sub-sample essentially reproduced those results presented in Figure 2, and in light of space constraints is not presented graphically. Specifically, fear was related to significantly higher perceived cost certainty (Std. Est. = .109, p < .01), severity (Std. Est. = .101, p < .05), and salience (Std. Est. = .123, p < .01), whereas anger was related to significantly lower cost severity (Std. Est. = −.110, p < .01) and salience (Std. Est. = −.141, p < .01). In addition, every other effect described previously in relation to Figure 2 was reproduced in the male-only model estimation, a result that would be expected given that 819 (81%) of the 1,013 in the total sample are males.

Results of the estimation of the final model from Figure 2, among the female sub-sample, revealed a few differences from the overall sample and male-only model estimations. It is unclear whether these reflect substantive differences, or simply result from having less statistical power among the smaller female sub-sample (n = 194). Regardless, this model continued to provide adequate fit to the underlying data, even with considerably fewer cases (ratio of χ² to degrees of freedom = 1.436, CN = 216, CFI = .928, RMSEA = .048). Fear was again related to marginally higher perceptions of cost certainty (Std. Est. = .160, p = .053) and significantly higher cost salience (Std. Est. = .198, p < .05). As was the case among males, fear did not directly impact hypothetical drunk driving likelihood, but anger did continue to exhibit a significant direct effect on drunk driving intentions (Std. Est. = .227, p < .05). Among the female sub-sample, anger did not impact the perception of any of the cost measures. Perceived salience of all reported costs was related to marginally lower drunk driving likelihood (Std. Est. = −.132, p = .077), yet the magnitude of the standardized effect was actually larger among females than it was in the male sample, where it had reached statistical significance (Std. Est. = −.092, p < .05).

Among females, cost certainty demonstrated a significant deterrent effect on drunk driving likelihood (Std. Est. = −.147, p < .05), while self-control was related to marginally lower drunk driving likelihood (Std. Est. = −.147, p = .059). Those who had driven drunk in the past without being caught also reported marginally higher intentions to drive drunk (Std. Est. = .148, p = .056). Finally, once again there was evidence for the experiential effect in that those who had driven drunk without being caught in the past reported significantly lower cost certainty perceptions (Std. Est. = −.170, p < .05).

As a means of summarizing the direct and indirect effects of each emotional state on hypothetical drunk driving likelihood, Table 3 contains the standardized total, direct, and indirect effects of each emotion scale, for the total sample, as well as among males and females separately. In each case, the entire effect of fear is an indirect effect through its influence on cost perceptions, while no direct effect emerges. For instance, in the overall sample, the total standardized effect for fear is −.032, while the direct effect is .000, such that 100% of the influence of fear is indirect. Conversely, for anger, the total standardized effect among the overall sample is .097, while the direct effect is .083, such that the indirect effect of anger (through altered cost perceptions) represents only about 14% of the total effect of anger on drunk driving intentions. This overall pattern is repeated in the male and female sub-samples. At least in these data, fear seems to only work through alteration of cost perceptions, whereas anger appears to work mostly through direct effects on drunk driving likelihood, exerting only a small indirect effect (about 14% of the total effect) through altered cost perceptions.

Table 3.

Direct and Indirect Standardized Effects of Emotions on Drunk Driving Intentions.

Sample	Variable	Standardized total effect on drunk driving	Standardized direct effect on drunk driving (% direct)	Standardized indirect effect on drunk driving (% indirect)
Total	Fear	−.032	.000 (0)	−.032 (100)
Total	Anger	.097	.083 (86)	.014 (14)
Males	Fear	−.027	.000 (0)	−.027 (100)
Males	Anger	.077	.064 (83)	.013 (17)
Females	Fear	−.050	.000 (0)	−.050 (100)
Females	Anger	.186	.167 (90)	.018 (10)

Discussion

The results presented in this study seem to support the hypothesis that states of fear and anger may have different influences on offending likelihood by altering the perception of potential costs of crime, particularly the salience of those costs in different ways. Although early research on affective valence suggested that negative emotions in general might lead individuals to engage in more deliberative, risk-averse decision-making processes, the results presented here suggest that two negatively valenced emotions (fear and anger) nonetheless have opposite effects on the perception of costs in a hypothetical offending situation. Specifically, even the modest levels of anger seen here decreased cost perceptions, and also directly influenced drunk driving intentions. Fear on the other hand, was related to increased perceptions of the severity and salience of costs, which then predicted lower intention to offend, but fear itself appeared to have no direct relationship with drunk driving intentions. These results occur controlling for the impact of other variables (e.g., self-control, the experiential effect) on cost perceptions, and appear generally consistent when each gender is examined separately (although the smaller sample of female offenders limited the statistical power of those analyses).

Little, if any, research in criminology has examined the impacts of fear and anger on cost perceptions, and as such the current study highlights the role that emotions play in criminal decision making. Importantly, it also does so among a diverse sample of serious offenders, rather than college students. Not only are emotional states considered as potential consequences of crime (as Bouffard et al., 2000, suggest), but a growing body of research supports Loewenstein’s (1996) suggestions about the attention-focusing effects of emotional/visceral states within the decision itself (i.e., on factors that are consistent with the experienced state). A number of previous studies have examined the role of sexual arousal, and to a lesser extent anger, demonstrating that these states can lead to increased attention to the possible benefits of each offense type (sexual coercion and fighting, respectively), thus facilitating these behaviors. The current study’s results extend our knowledge of the role of anger, by showing that it may also reduce the attention paid to costs of crime, while still exhibiting a direct effect on offending. These results also suggest that fear may invoke similar attention-focusing effects, though in the opposite direction, increasing the individual’s attention to the risks involved in hypothetical drunk driving.

Implications

Although this is but one study, this line of research holds the potential to improve both our theoretical understanding of criminal decision making, and the effectiveness of deterrence-based crime control programs. For instance, by explicitly modeling the relationship between emotional state and the perception of consequences within the decision, rational choice theory will likely be better able to explain behaviors that may otherwise seem “irrational,” but that have been influenced by the individual’s internal, emotional state. For instance, we now have evidence to suggest that angry individuals underestimate the costs of their misbehavior, whereas fearful people may give much more weight to costs. In relation to crime control efforts then, the current study’s results suggest that deterrence efforts may be made more effective if they attempt to instill states of fearfulness/anxiety in the would-be offender to increase cost perceptions at the crucial moment of decision.

Clarke and Homel (1996) have already developed a revised typology of situational crime prevention approaches that includes efforts to induce other negative emotional states, specifically shame and guilt, through the use of signage that reminds shoppers that “Shoplifting is Stealing,” for instance. It may be that these same kinds of efforts could be used to increase the individual’s level of fear and anxiety, which as Clarke and Homel (1996) describe would then “affect the situational calculus” of an offender (p. 25). A number of existing efforts, such as notifying shoppers that a store uses electronic surveillance, or relative to this study, billboards announcing that police will be conducting intoxicated driver roadblocks, could be used to increase would-be offenders’ current fear or anxiety. At the same time, existing research on “fear appeal” interventions suggests that such efforts are neither straightforward nor always effective (for a meta-analysis of these efforts within the public health arena, see Witte & Allen, 2000), so that additional research on how to design and implement these emotion-based strategies will likely also be needed. Future research should also examine how long any potential alteration of the individual’s emotional state may last, as this will have important effects on the magnitude of any additional deterrence produced by such interventions.

Beyond the potential to improve deterrence-based programs, the negative influence of anger on cost perceptions reinforces the value of efforts to reduce the would-be offender’s experience of this emotion, or at least to teach the individual to anticipate the impact that anger may have on future decisions. Loewenstein (1996) suggests that people in general underestimate the impact of future emotional states on their decisions. Offenders may have even greater difficulties in this area, due to their frequent lack of self-control and other skills to cope with emotional experiences like anger. At the same time, a growing body of research supports the use of anger management interventions (i.e., to recognize and cope with anger) as one of the most effective components of cognitive-behavioral treatment for offenders (see Landenberger & Lipsey, 2005). Likewise, relapse prevention programming that attempts to teach offenders to anticipate and plan for ways to deal with various “triggers” (anger and other negative emotions among them) has also been supported as effective (see Dowden, Antonowicz, & Andrews, 2003, for a meta-analytic example). The current study’s results underline the importance of these approaches by demonstrating that anger may in fact increase offending likelihood both directly and by reducing the perceived consequences of crime.

Limitations

While the results of the current study are an important first step in answering recent calls for increased attention to the role of choice and emotion in criminological theorizing, the current study is not without limitations. First, it uses a convenience sample of convicted felony offenders from a single Southwestern state, including a relatively smaller sample of female offenders. At the same time, none of the existing research on emotions and criminal decision making has used nationally representative samples, and in fact most of it has used relatively small convenience samples of undergraduate students. Thus, while studies are needed that examine the generality of these emotion effects on decision making among other samples (including the general population and non-incarcerated offenders), the current sample improves on what has been used in much of the existing research on emotions and criminal decision making, both in overall size and relevance (i.e., an actual offender sample).

Second, the current study did not experimentally manipulate levels of fear, but instead capitalized on existing, natural variation in current states of fearfulness and anger, using cross-sectional data. Previous research has shown that both experimentally manipulated and naturally occurring states of fear lead to increased risk perceptions (Lerner et al., 2003). At the same time, if experimental manipulations (and by extension, policy interventions) can be used to create even higher levels of fearfulness in future research, the relationships between fear and cost perceptions seen in this study may be even more pronounced. In addition, inmates were asked about their hypothetical likelihood of driving drunk while they are in a prison setting, which might inhibit honest responding. It is possible that this leads to some underreporting of actual future likelihood of drunk driving, although recall that 72% of the sample reported some probability of drunk driving. At the same time, about 84% said that they had done so in the past, so it is possible that this 12% point difference is reflective of some underreporting. If inmates do underreport their likelihood of driving drunk, this would tend to attenuate the relationship between current emotional state and offending likelihood, so again the significant effects seen in this study may underestimate the actual influence of emotions on decision making in real-world situations.

Finally, the current study uses a hypothetical intention design rather than actual behavioral outcomes. It is impractical and potentially unethical to assess the emotional state of an offender “on the street,” and then measure whether this influences his or her likelihood of actually committing a crime (including drunk driving). In fact, the use of the hypothetical design facilitates the assessment of current emotional state as it relates to a criminal decision, even if that decision is a hypothetical one. Thus, despite the limitations, numerous studies have demonstrated the validity of the hypothetical scenario design, in particular, for studying rational choice and perceptual deterrence theories (see Pogarsky, 2004, for a thorough discussion).

Despite these limitations, the current study, with its large sample of actual, serious offenders is the first in the field of criminology to examine how states of fearfulness may lead to increased deterrent effects by increasing the perception of crime’s costs, and likewise how anger may reduce the perception of costs and thus precipitate offending. These results also contribute to research in other fields (psychology in particular) examining the differential effects of various negative emotional states, by further demonstrating divergent effects of two negative emotions, fear and anger. A substantial body of research exists in fields such as psychology that links emotions such as fear and anxiety to risk-averse behaviors and decisions, yet little attention has been paid to this issue in criminology. As Nagin (2007) stated, “The long-term intellectual prosperity of criminology will depend on its strategically importing ideas and methods from outside its conventional disciplinary boundaries” (p. 269). The current study brings some of the interesting research being done in other social sciences to bear on an improved understanding of the role of emotions in criminal decision making and crime deterrence. Much work still needs to be done and hopefully this study will inspire additional criminological theorizing and research in this area. As both Nagin (2007) and Cullen (2011) remind us, the potential benefits for both basic understanding and effective intervention are immense.

Footnotes

Appendix

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Arbuckle

J. L.

(2011). IBM SPSS Amos 20 user’s guide. Chicago, IL: AMOS Development Corporation.

Ariely

Loewenstein

(2006). The heat of the moment: The effect of sexual arousal on sexual decision making. Journal of Behavioral Decision Making, 19, 87-98.

Bachman

Paternoster

Ward

(1992). The rationality of sexual offending: Testing a deterrence/rational choice conception of sexual assault. Law & Society Review, 26, 343-372.

Baron

R. A.

(1973). Threatened retaliation from the victim as an inhibitor of physical aggression. Journal of Research in Personality, 7, 103-115.

Baron

R. A.

(1974). Aggression as a function of victim’s pain cues, level of prior anger arousal and exposure to an aggressive model. Journal of Personality and Social Psychology, 29, 117-124.

Baron

R. A.

(1979). Effects of victim’s pain cues, victim’s race, and level of prior instigation upon physical aggression. Journal of Applied Social Psychology, 9, 103-114.

Barrett

(2007). Structural equation modeling: Adjudging model fit. Personality and Individual Differences, 42, 815-824.

Becker

(1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 169-217.

Blumstein

Cohen

Nagin

D. S.

(1978). Deterrence and incapacitation: Estimating the effects of criminal sanctions on crime rates. Washington, DC: National Academy of Sciences.

10.

Bouffard

J. A.

(2002a). The influence of emotion on rational decision making in sexual aggression. Journal of Criminal Justice, 30, 121-134.

11.

Bouffard

J. A.

(2002b). Methodological and theoretical implications of using subject-generated consequences in tests of rational choice theory. Justice Quarterly, 19, 747-771.

12.

Bouffard

J. A.

(2011). In the heat of the moment? Mediating vs. moderating relationships between sexual arousal and perceived sanctions. Journal of Crime & Justice, 34, 24-44.

13.

Bouffard

J. A.

Exum

M. L.

Paternoster

(2000). Whither the beast? The role of emotions in a rational choice theory of crime. In Simpson

S. S.

(Ed.), Of crime and criminality: Applying criminological theory in everyday life (pp. 159-178). Thousand Oaks, CA: Pine Forge Press.

14.

Carmichael

Piquero

A. R.

(2004). Sanctions, perceived anger and criminal offending. Journal of Quantitative Criminology, 20, 371-393.

15.

Chen

Curran

P. J.

Bollen

K. A.

Kirby

Paxton

(2008). An empirical evaluation of the use of fixed cutoff points in RMSEA test statistic in structural equation models. Socialogical Methods & Research, 36, 462-494.

16.

Clarke

R. V.

Felson

(1993). Introduction: Criminology, routine activity, and rational choice. In Clarke

R. V.

Felson

(Eds.), Routine activity and rational choice (pp. 1-13). London, England: Transaction.

17.

Clarke

R. V.

Homel

(1996). A revised classification of situational crime prevention techniques. In Lab

(Ed.), Crime prevention at the crossroads (pp. 17-27). Cincinnati, OH: Anderson.

18.

Cornish

D. B.

Clarke

R. V.

(1987). Understanding crime displacement: An application of rational choice theory. Criminology, 25, 933-947.

19.

Cullen

F. T.

(2011). Beyond adolescence-limited criminology: Choosing our future—The American Society of Criminology 2010 Sutherland address. Criminology, 49, 287-330.

20.

Cusson

(1993). Situational deterrence: Fear during the criminal event. In Clarke

(Ed.), Crime prevention studies (Vol. 1, pp. 55-68). Monsey, NY: Criminal Justice Press.

21.

Damasio

A. R.

(1994). Descartes’ error: Emotion, reason, and the human brain. New York, NY: Putnam.

22.

Dowden

Antonowicz

Andrews

D. A.

(2003). The effectiveness of relapse prevention with offenders: A meta-analysis. International Journal of Offender Therapy and Comparative Criminology, 47, 516-528.

23.

Exum

M. L.

(2002). The application and robustness of the rational choice perspective in the study of intoxicated and angry intentions to aggress. Criminology, 40, 933-967.

24.

Forgas

J. P.

(1989). Mood effects on decision making strategies. Australian Journal of Psychology, 41, 197-214.

25.

Forgas

J. P.

(1995). Mood and judgment: The affect infusion model (AIM). Psychological Bulletin, 117, 39-66.

26.

Frijda

N. H.

(1969). Recognition of emotion. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 4, pp. 167-223). New York, NY: Academic Press.

27.

Grasmick

H. G.

Bursik

R. J.

(1990). Conscience, significant others and rational choice: Extending the deterrence model. Law & Society Review, 24, 837-861.

28.

Grasmick

H. G.

Tittle

C. R.

Bursik

R. J.

Jr. Arneklev

B. K.

(1993). Testing the core empirical implications of Gottfredson and Hirschi’s general theory of crime. Journal of Research in Crime and Delinquency, 30, 5-29.

29.

Green

(1989). Measures of illegal behavior in individual-level deterrence research. Journal of Research in Crime & Delinquency, 26, 253-275.

30.

Higgins

G. E.

Wilson

A. L.

Fell

B. D.

(2005). An application of deterrence theory to software piracy. Journal of Criminal Justice and Popular Culture, 12, 166-184.

31.

Johnson

E. J.

Tversky

(1983). Affect, generalization and the perception of risk. Journal of Personality and Social Psychology, 45, 20-31.

32.

Karstedt

Loader

Strang

(2011). Emotions, crime and justice. Oxford, UK: Hart.

33.

Landenberger

N. A.

Lipsey

M. W.

(2005). The positive effect of cognitive-behavioral programs for offenders: A meta-analysis of factors associated with effective treatment. Journal of Experimental Criminology, 1, 451-476.

34.

Lerner

J. S.

Gonzalez

R. M.

Small

D. A.

Fischhoff

(2003). Effects of fear and anger on perceived risk of terrorism: A national field experiment. Psychological Science, 14, 144-150.

35.

Lerner

J. S.

Keltner

(2000). Beyond valence: Toward a model of emotion-specific influences on judgment and choice. Cognition & Emotion, 14, 473-493.

36.

Lerner

J. S.

Keltner

(2001). Fear, anger and risk. Journal of Personality and Social Psychology, 81, 146-159.

37.

Loewenstein

(1996). Out of control: Visceral influences on behavior. Organizational Behavior and Human Decision Process, 65, 272-292.

38.

Loewenstein

Lerner

J. S.

(2003). The role of affect in decision making. In Davidson

Sherer

Goldsmith

(Eds.), Handbook of affective science (pp. 619-642). New York, NY: Oxford University Press.

39.

Loewenstein

Nagin

D. S.

Paternoster

(1997). The effect of sexual arousal on expectations of sexual forcefulness. Journal of Research in Crime & Delinquency, 34, 443-473.

40.

Nagin

D. S.

(1998). Criminal deterrence research at the outset of the twenty-first century. In Morris

Tonry

(Eds.), Crime and justice: A review of research (Vol. 23, pp. 1-42). Chicago, IL: University of Chicago Press.

41.

Nagin

D. S.

(2007). Moving choice to center stage in criminological research and theory: The American Society of Criminology 2006 Sutherland address. Criminology, 45, 259-272.

42.

Nagin

D. S.

Paternoster

(1993). Enduring individual differences and rational choice theories of crime. Law & Society Review, 27, 467-496.

43.

Nagin

D. S.

Paternoster

(1994). Personal capital and social control: The deterrence implications of a theory of individual differences in criminal offending. Criminology, 32, 581-606.

44.

Nagin

D. S.

Pogarsky

(2001). Integrating celerity, impulsivity, and extralegal sanction threats into a model of general deterrence: Theory and evidence. Criminology, 39, 865-889.

45.

Novaco

(2011). Anger dysregulation: Driver of violent offending. Journal of Forensic Psychiatry and Psychology, 22, 650-668.

46.

Paternoster

(1987). The deterrent effect of perceived certainty and severity of punishment: A review of the evidence and issues. Justice Quarterly, 4, 173-217.

47.

Paternoster

Saltzman

L. E.

Waldo

G. P.

Chiricos

T. G.

(1983). Perceived risk and social control. Law & Society Review, 17, 457-479.

48.

Peters

Vastfjall

Garling

Slovic

(2006). Affect and decision making: A “hot” topic. Journal of Behavioral Decision Making, 19, 79-85.

49.

Piquero

A. R.

Bouffard

J. A.

(2007). Something old, something new: A preliminary investigation of Hirschi’s redefined self-control. Justice Quarterly, 24, 1-27.

50.

Piquero

A. R.

Tibbetts

S. G.

(1996). Specifying the direct and indirect effects of low self-control and situational factors in offenders’ decision-making: Toward a more complete model of rational offending. Justice Quarterly, 13, 481-510.

51.

Pogarsky

(2002). Identifying “deterrable” offenders: Implications for research on deterrence. Justice Quarterly, 19, 431-452.

52.

Pogarsky

(2004). Projected offending and contemporaneous rule-violation: Implications for heterotypic continuity. Criminology, 42, 111-136.

53.

Pratt

T. C.

Cullen

F. T.

(2005). Assessing macro-level predictors and theories of crime: A meta-analysis. In Tonry

(Ed.), Crime and justice: A review of research (Vol. 32, pp. 373-450). Chicago, IL: University of Chicago Press.

54.

Schwarz

Clore

G. L.

(2007). Feelings and phenomenal experiences. In Kruglanski

Higgins

E. T.

(Eds.), Social psychology: handbook of basic principles (pp. 385-407). New York, NY: Guilford Press.

55.

Smith

C. A.

Ellsworth

P. C.

(1985). Patterns of cognitive appraisal in emotion. Journal of Personality and Social Psychology, 48, 813-838.

56.

Stafford

M. C.

Warr

(1993). A reconceptualization of general and specific deterrence. Journal of Research in Crime & Delinquency, 30, 123-135.

57.

Tibbetts

S. G.

(2013). Traits and states of self-conscious emotions in criminal decision making. In VanGelder

J. L.

Elffers

Reynaud

Nagin

(Eds.), Affect and cognition in criminal decision making (pp. 221-238). London, England: Routledge.

58.

Tibbetts

S. G.

Myers

D. L.

(1999). Low self-control, rational choice, and student test cheating. American Journal of Criminal Justice, 23, 179-200.

59.

Tiedens

L. Z.

Linton

(2001). Judgment under emotional certainty and uncertainty: The effects of specific emotions on information processing. Journal of Personality and Social Psychology, 81, 973-988.

60.

Watson

Clark

L. A.

Tellegen

(1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063-1070.

61.

Witte

Allen

(2000). A meta-analysis of fear appeals: Implications for effective public health campaigns. Health Education & Behavior, 27, 591-615.

62.

Zillman

(1979). Hostility and aggression. Hillsdale, NJ: Erlbaum.