II. History-Sensitive Sentencing

I shall refer to the practice of increasing or decreasing a prospective sentence based on criminal history, or lack thereof, as ‘history-sensitive sentencing’ (HSS). HSS allows criminal history, or lack thereof, to function as aggravating or mitigating factors, respectively. Sentencing authorities practice HSS if and only if they give a repeat offender a heavier sentence, on account of the criminal record, than they would impose on an otherwise identical offender with a less extensive record.

We can classify sentencing regimes by how much HSS, if any, they prescribe and in what forms. The simplest regime, flat-rate sentencing, involves no HSS whatsoever. Courts sentence based exclusively on features of the charged offense.Footnote ⁹ If we plot the sentencing function with criminal history on the y-axis and sentences on the x-axis, its slope is flat.

Next, consider a sentencing function that slopes upward until the n _th offense and then flattens. Sentences plateau or ‘cap’ at a pre-established maximum. This is known as ‘progressive loss of mitigation’ (PLM).Footnote ¹⁰ A special case of capping is a function that slopes upward immediately after the first offense and flattens thereafter—the ‘first-offender discount’ (FOD).

The final category of HSS—usually the most severe, in practice—is known as ‘prior-record enhancement’ (PRE) or the ‘recidivist premium’ (RP). Under the RP, a repeat offender receives progressively heavier sentences, possibly with a cap on total years served.

Different regimes implement different forms of HSS and use different legal mechanisms.Footnote ¹¹ In some jurisdictions, sentencing authorities are authorized to take criminal history into account on a discretionary basis.Footnote ¹² In the USA, such discretion is now usually restricted by statute and/or guided by published guidelines (mandatory or advisory). Some criminal statutes specify prior convictions as aggravators or define separate sentencing ranges for recidivists. Sentencing guidelines often permit the RP. The USA has, overall, the world’s most severe prior-record enhancements.Footnote ¹³

A criminal record includes, inter alia, arrests, indictments, convictions, sentences originally imposed, and months actually served. The federal guidelines assign convicts to one of six Criminal History Categories based on ‘points’ for different entries, subject to departures.Footnote ¹⁴ They also single out ‘career offenders’ for special treatment.Footnote ¹⁵ Guidelines in nineteen US states use grids or worksheets, assigning points or point ranges.Footnote ¹⁶

I make some simplifying assumptions in this article because my interest is justifications for HSS, simpliciter. I use ‘number of prior convictions’ as a proxy for the record as a whole. This assumption is neither a reflection of positive law nor a prescription. If we find a sound rationale for HSS, we can then pose questions about which entries on the record to treat as enhancers and how much significance to give them.

Also for simplicity, I assume sequences of identical crimes. Real records typically contain diverse offenses of varying severity. Graded versions of HSS proportion enhancements to severity. Ungraded versions ignore severity, considering only the number of convictions. Some three-strikes laws, for example, mandate a relatively ungraded premium, treating as strikes all felonies, regardless of severity.Footnote ¹⁷ Some states include misdemeanors as sentence aggravators. Some allow convictions that are many decades old, including juvenile convictions, to greatly enhance sentences.

III. Forward-looking Considerations

A natural thought is that repeat offenders have shown themselves to be dangerous, incorrigible people who should be kept in prison to protect the public. Indeed, arguments for HSS often cite forward-looking reasons, chiefly those pertaining to crime control.Footnote ¹⁸ These include incapacitation (just mentioned), along with special deterrence, general deterrence, and (arguably) rehabilitation. If repeat offenders are more likely to reoffend than are first offenders, there may be arguments for HSS from incapacitation and special deterrence.Footnote ¹⁹

Some commentators also make forward-looking arguments against HSS, especially the extreme RP employed in many of the United States. Long-term incarceration strains state budgets, destroys lives, and damages communities.Footnote ²⁰

This article mostly ignores forward-looking arguments (pro and con) for three reasons. First, my intended readers are philosophers and legal scholars engaged with the philosophy of criminal law. In these circles, forward-looking arguments are recognized, but never get the last word. Justice is considered at least as important as crime control.

Secondly, existing sources of law imply rationales for HSS that are at least partly retributivist, using the language of desert, culpability, and blame: “[w]hether one accepts the retributive justifications or not, the reality is that many statutes and most guidelines schemes reflect or actively promote the view that repeat offenders are more blameworthy and hence deserving of more punishment.”Footnote ²¹

The US federal sentencing guidelines, for example, mention deterrence and incapacitation, but also state, “A defendant with a record of prior criminal behavior is more culpable than a first offender and thus deserving of greater punishment.”Footnote ²²

Thirdly, members of the general public support HSS at least in part because they agree that recidivists, as such, deserve heavier sentences.Footnote ²³ Nevertheless, reconciling HSS with a developed retributivist theory has proved challenging.Footnote ²⁴ In §§IV-IX, I consider an argument for HSS aiming for such reconciliation.

IV. Retributivist Sentencing

There are many retributivist principles in the literature, but all entail that what the convict deserves is relevant to sentence justification.Footnote ²⁵ Retributive rationales for various forms of HSS have been debated in recent years.Footnote ²⁶ Some writers have concluded that no form of HSS is, in fact, compatible with a defensible retributivism, such that they must choose between retributivism and HSS.Footnote ²⁷ Most choose retributivism. As noted, HSS is part of the penal status quo worldwide, so their position entails that the status quo is unjust.

Other retributivists support HSS, but only in its weaker forms (typically the FOD or PLM).Footnote ²⁸ Even retributivists who favor the RP overwhelmingly reject its more extreme forms.Footnote ²⁹ Most retributivists oppose ungraded enhancements, such as three-strikes laws (see §II). Some insist that the “primary determinant” of a sentence should be the most recent conviction, not the convict’s history, which entails that sentences should not reflect judgments of “overall character” or “lifetime conduct.”Footnote ³⁰ The primacy of the present conviction also implies that the maximum RP should remain low, in relative terms, such that the premium’s length should never exceed the sentence imposed for the same crime as a first offense, or some small multiple thereof.Footnote ³¹ Others insist that the maximum RP should remain low, in absolute terms, which has been defined variously as “[no] more than one-half of the total sentence”Footnote ³² or “less than five additional years” or “substantially less than life without parole.”Footnote ³³ However, more severe versions of the RP are implemented in many jurisdictions.Footnote ³⁴ These policies are seen as unjust even by the retributivists most supportive of HSS.

V. History, Desert, and Sentencing

The literature contains several arguments for HSS that purport to be compatible with familiar retributivist principles.Footnote ³⁵ Below is the bare bones of such an argument, which I shall develop and evaluate.

Therefore,

Many find premise 7 (HRD) to be intuitive, but it remains controversial.Footnote ³⁷ Some insist that the punishment should ‘fit the crime,’ not the record.Footnote ³⁸ A convict’s history, in itself, seems to have no evidentiary relevance to his desert for subsequent crimes. Some complain that the RP “punish[es] a person twice for the same offense.”Footnote ³⁹ Retributivists are “deeply divided on whether, why, and to what extent a repeat offender deserves more punishment.”Footnote ⁴⁰ As Frase observes, “Desert theory is very underdeveloped when it comes to providing principles and practical decision rules that can both justify and limit prior-conviction enhancements.”Footnote ⁴¹ Even retributivists who accept HRD disagree with one another about how to defend it. The literature contains several arguments for it, none decisive.Footnote ⁴²

VI. Character and Culpability

An intuitive way to support HRD involves inferring from criminal history, to character, to desert. Consider, first, the inference from character to desert. Could bad character, itself, warrant punishment?Footnote ⁴³ It might seem otherwise. The act requirement requires bad acts (or omissions) for criminal conviction, precluding punishment for bad character, absent action.Footnote ⁴⁴ Nevertheless, the state could, in theory, treat certain bad character traits as enduring, universal aggravators of negative desert for bad acts. For as long as a convict retained such traits, they would deserve more punishment for any crime they committed than would someone without such traits (if convicted of the same offense). Conversely, the state could treat good character traits as enduring, universal mitigators of negative desert.

Treating bad character as a basis for enduring, universal penalty enhancement is a conceivable policy. However, there is a conclusive, logical reason why evidence of bad character, if based on criminal history, cannot justify such a policy. For conviction, the prosecution must produce new evidence to prove that the defendant is more likely to be guilty than was previously established (very likely, in fact). Similarly, a convict should be sentenced to additional punishment for possessing a bad character trait only if the prosecution proves that the convict is more likely to possess the trait, or possesses it to a greater degree, than was previously established. Criminal history is, by definition, evidence available to the state before the trial began. Additional punishment is justified only if evidence presented at trial establishes that the convict is more likely to possess the bad trait, or possesses it to a greater degree, than the state previously knew. And that is possible only if the trait was a cause of new evidence presented. Character traits that are causally irrelevant to the sentencing offense are irrelevant to any argument for HSS that proceeds from criminal history, to character, to desert.

For this reason, arguments with this structure cannot appeal to traits, simpliciter, but only to traits for which the sentencing offense provides new evidence. These, by definition, are traits that figure in good causal explanations of the sentencing offense. This is to say that the traits must bear on desert in virtue of their relevance to the convict’s degree of culpability for the sentencing offense, also called scalar culpability (i.e., how culpable they are, not whether they are culpable). Unsurprisingly, according to two leading experts on HSS, “most desert-based theories of prior record enhancement have been based on mitigated or enhanced offender culpability.”Footnote ⁴⁵

Define a ‘culpability-relevant character trait’ as one that is relevant to an agent’s scalar culpability for their (independently blameworthy) actions. Using that definition, I reconstruct the argument as follows:

Therefore,

All retributivists treat culpability as a desert input,Footnote ⁴⁶ so all would accept premise 2. They should be interested in whether criminal history can constitute evidence of enhanced culpability.

Different theories of scalar culpability recognize different culpability bases. All theories assign weight to the nature of one’s conductFootnote ⁴⁷ and mental states (e.g., beliefs, motives, intentions). Some also consider relevant the nature and extent of any harm caused, although this factor is controversial.Footnote ⁴⁸ Premise 3 states that certain character traits can also be relevant—a ‘character-sensitive theory of culpability’ (CSTC).Footnote ⁴⁹

Character judgments abound in folk morality, law, and philosophical ethics.Footnote ⁵⁰ Many of the traits of folk psychology are intuitively relevant to scalar culpability. Plausible examples include greed, sadism, lechery, cowardice, dishonesty, disloyalty, treachery, laziness, vindictiveness, selfishness, callousness, intemperance, maybe even gluttony (I hope not). However, there is no satisfactory theory of the relevance of traits to scalar culpability. My question concerns not the appropriateness of blaming someone for their traits, per se, but of judging them culpable to a greater (or lesser) degree for conduct that is connected in certain ways to certain traits of theirs. I present a theory of traits that can enhance culpability for otherwise blameworthy acts such that enhanced censure, including enhanced punishment, might be deserved and justified, in principle. I argue that such a theory is coherent and intuitively plausible, although its usefulness in justifying HSS is debatable.

In §§IX-XIII, I shall critically evaluate arguments for premise 6 (HRC) that rely explicitly upon a CSTC. However, the literature also contains important arguments for HRC that do not rely explicitly upon character ascriptions as a middle term. I postpone their discussion until §XIV.

For now, I shall insert propositions 2-6 into the generic argument from §V, yielding the character argument for the relevance of criminal history to retributive sentencing (the ‘Character Argument’):

Therefore,

Therefore,

Therefore,

Premises 1-3 are a priori; 4 and 5 are a posteriori. My remaining project in this article is to understand and probe the Character Argument (CA).

VII. The Nature of Character Traits

Premise 3 of the CA states that certain character traits are relevant to scalar culpability. To evaluate this premise, we must address some questions about character traits, simpliciter. Folk psychology is my point of departure, although not my final destination. In §§VII-VIII, I describe a folk-psychological conception of character traits. Whether such traits actually exist is another question—an empirical question—to be raised but not answered in §§X-XII.

VIII. Aggravation and Mitigation

The model of personality traits just presented aims to capture how folk psychology conceives them: as dispositional, motivationally causal, stable over time, consistent across situations, and potentially manifested in inclination and action. We can now address the relevance of traits to scalar culpability. In doing so, we narrow our focus to character traits—the subset of personality traits directly relevant to moral appraisal of agents.Footnote ⁶⁴

I propose a specification of a CSTC comprising two principles. Each principle includes two relevance conditions, individually necessary and jointly sufficient. The first condition in each case links agents with traits. The second links traits with acts. The principles are:

Aggravation and Mitigation are ceteris paribus principles. To say that a trait aggravates an agent’s culpability for an act is to say that, holding constant their other culpability-relevant attributes (e.g., nature of act, content of inclination set, any harm caused, etc.), the agent is more culpable for the act if the trait is a cause of the operative inclination set and manifested thereby.Footnote ⁶⁶ An agent’s inclination set is culpability-aggravating if they are more culpable than they would otherwise be for a blameworthy act if the set is a cause thereof and manifested thereby. An inclination set is more culpability-aggravating, ceteris paribus, if a culpability-aggravating trait is a cause of and manifested by it.

Which inclination sets aggravate culpability? One could simply use folk judgments about positive and negative motives and desires, but for purpose of discussion I suggest the following analysis: an inclination set aggravates culpability if and only if acting from it manifests insufficient concern for the legally or morally protected interests of others.Footnote ⁶⁷

Aggravation and Mitigation reference ‘agential implication’—a two-place relation between agent and trait. A trait that is not agency-implicating cannot serve the CA, even if it meets all other conditions for culpability-relevance. A trait is agency-implicating for agents if and only if it bears on our normative appraisal of them. However, philosophers disagree about the conditions under which traits are agency-implicating. Some claim, for example, that a trait is agency-implicating only if it is part of the agent’s deep self or real self.Footnote ⁶⁸ Others deny this.Footnote ⁶⁹ Some claim, while others deny, that the agent must be morally responsible for the trait, itself.Footnote ⁷⁰ Whether a trait is agency-implicating, and therefore useful to the CA, depends on the correct theory of agential implication and whether the trait meets its conditions (see §XII.D for further discussion).

We now have a provisional account of the character traits of folk psychology and principles linking them to culpability: Aggravation and Mitigation. These imply that criminals are more culpable for their crimes, ceteris paribus, if they possess agency-implicating traits that are causes of and manifested by culpability-aggravating inclination sets that are causes of their crimes. They are less culpable if either condition is not met, although they remain culpable for crimes that reflect their wrongful choices (see §IX.C). They are still less culpable, ceteris paribus, if their crimes either 1) manifest and are caused by culpability-mitigating traits; or 2) contra-manifest and are impeded by culpability-mitigating traits.

I turn next to some philosophical objections to traits as culpability-aggravating.

IX. Objections to Traits as Aggravating Culpability

X. The Nature and Incidence of Culpability-Relevant Traits

The CA may be coherent and inferentially valid, with plausible a priori premises, but its soundness is far from demonstrated. The remainder of this article is more critical of the CA, shifting attention to a posteriori premises, beginning with premise 4, which states that many convicts have culpability-relevant character traits. The following objection directly challenges premise 4 as to culpability-aggravating traits:Footnote ⁸³

Therefore,

Therefore,

This conclusion (v) contradicts premise 4 of the CA as to culpability-aggravating traits. However, premise (ii) of the No-Such-Traits Objection (NSTO) contradicts folk psychology, which posits a diverse set of robust character traits (see §VII). It is intuitively plausible that some such traits manifest in blameworthy acts. Therefore, premise (ii) is counterintuitive.

Nevertheless, premise (ii) must be entertained, because the incidence of robust traits is empirically controversial. For over a century, experimental psychologists have mounted sweeping challenges to the folk psychology of personality traits. Some experiments suggest that robust traits are not widespread in the general population and that non-trait factors, rather than traits, are the dominant causes of behavior. The emerging psychological hypotheses are known as ‘situationism.’Footnote ⁸⁴ I shall refer to the cited experiments as the ‘character experiments.’ The more famous of these include Stanley Milgram’s obedience experiment at Yale, the Stanford prison experiment, and the Good Samaritan experiments conducted in Princeton, New Jersey. There are hundreds of others in the situationist literature.Footnote ⁸⁵ Many involved randomized, controlled, experimental manipulation of situations, ensuring that the subject’s encountering the situation cannot itself be attributed to any stable traits (e.g., traits that might lead one to seek out such situations). The typical finding reveals that a particular type of behavior varies more for one agent across different situation-types than it does across different agents facing one situation-type. In such cases it appears that the situation-type, not individual personality traits, is the dominant determinant of the behavior.

One conclusion drawn by psychologists is that slight variations in the agent’s situation (e.g., whether they are in a hurry or have recently received some good news or have been instructed to act) make a surprisingly strong difference to their likelihood of behaving as a given trait would indicate under the circumstances. The experimental subjects’ behavior tracks variations that everyone, including the subjects, themselves, would consider morally irrelevant.Footnote ⁸⁶ Much of our behavior appears susceptible to influences we would not reflectively endorse as good reasons for action. Traits appear to play much less of a role than folk psychology predicts.

Some psychologists and philosophers have interpreted these results as posing a serious challenge to the folk psychology of character. Some conclude that robust traits are very rare or non-existent in the general population.Footnote ⁸⁷ Arguments from empirical inadequacy defy quick summary, but cannot be quickly dismissed, as they are based on extensive study of the scientific literature.

XI. Beyond Folk Psychology

Debate continues about the lessons of the character experiments. Their methodologies have been questioned, as have interpretations of the results as supporting skepticism about robust traits.Footnote ⁸⁸ This is not the place to survey that debate. I focus instead upon a complementary tactic for rebutting the NSTO. This tactic challenges premise (i), which states that a trait type is culpability-relevant only if robust and manifested in blameworthy/praiseworthy acts. Suppose there is an alternative conception of traits, one that survives the character experiments. If traits so conceived exist and are relevant to culpability, premise (i) is false.

One might find such a conception in recent defenses of virtue ethics against empirical inadequacy critiques. Critics such as John Doris argue that virtue ethics presupposes the folk-psychological conception of character traits and that the character experiments reveal such traits to be so rare that virtue ethics is not a realistic prescriptive theory for human beings.Footnote ⁸⁹ Virtue ethicists have responded by clarifying and refining their philosophical commitments, frequently distancing themselves from folk psychology in the process.Footnote ⁹⁰ They suggest alternative conceptions of traits that could prove useful in challenging premise (i) of the NSTO. Several virtue ethicists emphasize that traits do not invariably produce behavior. Rather, they generate pro tanto inclinations that vary along the dimension of strength.Footnote ⁹¹

Another conceptual refinement advocated by some virtue ethicists qualifies the dimension of cross-situational consistency introduced in §VII.B. The width of a trait reflects the diversity of situations in which it manifests. According to Doris, both folk psychology and virtue ethics posit wide traits. However, some virtue ethicists reply that they are not (and, perhaps, were never) committed to wide traits, folk psychology notwithstanding.

The question remains whether narrower traits are culpability-relevant to the extent required by the CA. Let us define ‘revisionary traits’ as any that differ from folk psychology and are weak enough and/or narrow enough to survive the character experiments. Notwithstanding the experiments, certain revisionary traits might be widely instantiated in the general population. Even critics of robust traits believe in certain revisionary traits. Doris appears to endorse revisionary traits that are relatively strong, but very narrow.Footnote ⁹² If such traits are widespread in the general population, it is plausible that criminals also have them. Perhaps it is even the case that, as compared to law-abiding citizens, the average criminal has more numerous revisionary traits that are negative and fewer that are positive.

Assume, arguendo, that certain negative, revisionary traits are pervasive in criminal populations. The next question is whether such traits are culpability-aggravating. We must proceed with caution. Folk moral judgments presuppose folk psychology, not empirical psychology. Folk judgments of culpability-relevance refer to robust traits, not revisionary ones. As we revise our conception of traits away from folk psychology, we compromise our ability to rely on associated folk-moral judgments. Whether or not folk psychology posits revisionary traits alongside robust ones, it is not obvious that folk morality sees revisionary traits as culpability-relevant at all, much less to the extent the CA requires. Folk judgments of relative culpability might be implicitly calibrated to robust traits, not revisionary ones.

In a similar vein, some legal scholars have worried that revisionary conceptions of character threaten something more fundamental than the CA—the very idea of criminal responsibility.Footnote ⁹³ Anders Kaye wonders if “the more modest character traits sufficient … for virtue theory can adequately sustain attributions of responsibility.”Footnote ⁹⁴ Donald Dripps worries that “[a] virtue theory that did describe traits defined narrowly enough to be recognized in the psychological research … would have to account for an unmanageably large number of predispositions.”Footnote ⁹⁵

I shall not argue that revisionary traits are incapable of supporting legal and moral responsibility. My question is more modest: can they serve the CA? The Aggravation Principle entails that a trait aggravates culpability only if it is motivationally causal, manifested in blameworthy action, and agency-implicating. Therefore, the CA requires traits that are manifested in criminal acts (which implies that they are crime-causing, ‘criminogenic’) and agency-implicating. Such traits must be relatively stable, by definition, and instantiated in the criminal population. They must be narrow and/or weak enough to survive the character experiments, but also wide enough and strong enough to be both manifested in criminal acts and agency-implicating. The widest, strongest traits are agency-implicating and plausibly manifested, but the character experiments suggest they are not empirically adequate. The narrowest, weakest traits may be empirically adequate, but it remains to be established that they are sufficiently manifested in crime.

With these constraints in mind, I can now post a ‘job description’ for traits that could serve as the culpability-aggravators driving the CA. They should be 1) relatively stable; 2) at least medium-width; 3) at least moderately strong; 4) agency-implicating; and 5) significantly crime-manifested, which implies 6) significantly criminogenic.

XII. Candidate Traits

As we search for traits with the aforementioned qualifications, criminologists are natural expert witnesses. This section begins that consultation, in a speculative and rudimentary way. For illustrative purposes, I focus on a trait for which there is actual evidence of criminogenesis. I proceed to evaluate it against the rest of the job description. My conclusions are mixed.

Some theories of criminogenesis assign personality traits no role or only a minimal role.Footnote ⁹⁶ Such theories complement the character experiments. Just as folk psychology posits causal personality traits in the general population, it posits criminogenic traits in the criminal population. The character experiments suggest that folk psychology is unreliable with respect to character in general. Criminogenic theories that deemphasize personality traits imply that folk psychology is similarly unreliable with respect to criminal character.Footnote ⁹⁷ If personality traits are not criminogenic, they cannot be relevant to culpability under a CSTC. Therefore, the defender of the CA must hope that criminologists on the opposing side of the debate—those positing criminogenic personality traits—are closer to the truth.

I cannot review the vast criminogenesis literature, but use an illustrative example: a theory of criminogenesis that might support the existence of a personality trait suited to the purposes of the CA. This is the self-control theory (SCT), a widely cited theory of criminogenesis that is central to Michael Gottfredson and Travis Hirschi’s controversial general theory of crime. Gottfredson and Hirschi posit a “stable construct,” a single, unidimensional, latent trait—degree of self-control—and assert that it is “the individual characteristic relevant to the commission of criminal acts.”Footnote ⁹⁸ Some of their theses with significance for my purposes are as follows:

• The main cause of crime is low self-control.

• Someone with high self-control is substantially less likely to commit crimes at every stage of his life.

• Someone is more likely to commit crimes the less self-control he has.

• Someone with low self-control will not inevitably commit crimes.

• Low self-control has many manifestations, only some of which are criminal.Footnote ⁹⁹

It is crucial to distinguish colloquial meanings of ‘self-control’ from this criminological term of art. In a colloquial sense of ‘self-control,’ any properly convicted criminal lacked self-control, by definition, when they offended, because they did not ‘control’ themself. In this sense, it is tautological that convicts lacked self-control (past tense).

In criminology, by contrast, ‘self-control’ denotes a particular empirical construct: a stable disposition that can be given an operational definition.Footnote ¹⁰⁰ Using such a definition, the proposition that criminals have less ‘self-control’ than non-criminals is a falsifiable empirical hypothesis about stable dispositions, not a tautology about past criminal activity.

According to Gottfredson and Hirschi, low self-control (LSC) is associated with the following attributes, inter alia:

• Present oriented (tendency to respond to tangible stimuli in the immediate environment)

• Absence of diligence, tenacity, or persistence in a course of action

• Little interest in and unprepared for long-term occupational pursuits

• Adventuresome, active, physical

• Self-centered, indifferent, or insensitive to suffering and needs of others

• Temper (minimal tolerance for frustration)

• Possibly limited cognitive skills

• Sociable.Footnote ¹⁰¹

They claim that these attributes, which are at least indirectly observable, are correlated with one another and collectively constitute evidence of an underlying trait they call ‘low self-control.’ I can now address the relationship between self-control, as the SCT understands it, and my project. How, if at all, does degree of self-control map onto personality traits? Is LSC, itself, a personality trait or an effect or aspect of one? Is there one trait (or set thereof) that correlates with high self-control and another with low self-control? Assume, for the purpose of this section, that LSC is non-trivially instantiated in the general population and that there is, indeed, a personality trait that correlates with it. I shall call it a ‘trait of low self-control’ (TLSC), to distinguish it conceptually from LSC, simpliciter, which might or might not satisfy my definition of a trait.

My next question is whether a TLSC is culpability-aggravating and, if so, to the extent required by the CA. As discussed in §XI, viable candidates will be 1) relatively stable; 2) at least medium-width; 3) at least moderately strong; 4) agency-implicating; and 5) significantly crime-manifested, which implies 6) significantly criminogenic. In a preliminary way, I shall assess a TLSC against these criteria.

XIII. Traits and Criminal History

The inconclusive lesson of the previous section is that a TLSC—an initially promising candidate—might or might not serve the purposes of the CA. Of course, other revisionary traits could be described and assessed as alternatives. For the sake of this section, let us assume that some traits—perhaps a TLSC, perhaps others—are culpability-aggravating and possessed by many convicts. Premise 5 of the CA rests on the answer to yet another empirical question, which concerns how traits are distributed across convict populations such as this one:

For any version of HSS, the CA can support it only if it is possible, in principle, to sort a subpopulation of convicts, such as the Florida Burglars, by scalar culpability into at least as many distinct clusters as the prescribed number of sentence enhancement tiers.Footnote ¹⁰⁹ The FOD has two tiers, so it requires two clusters, representing the degrees of culpability ascribable to first offenders versus repeat offenders. PLM, by contrast, has at least three tiers and requires at least three clusters. Finally, the RP might require more than three clusters to match its tiers. Consider the following definition:

Imagine bimodality holds for the Florida Burglars: they bifurcate into clusters based on culpability-relevant traits. I shall use the FOD as my representative version of HSS. The FOD imposes a longer sentence on every repeat-offender within the Florida Burglars than on every first offender. These are ideal sentences according to premise 1 of the CA (§VI) only if the higher-culpability cluster consists entirely of repeat offenders and the lower-culpability cluster entirely first offenders. In other words, every first offender has more culpability-aggravating traits and fewer culpability-mitigating traits than does every repeat offender.

This is unrealistic, of course. The correlation between recidivism and bad character is surely imperfect. But the CA does not require perfect bimodality. The correlation just needs to be positive and sufficiently strong. Sentencing standards and practices are always crude, relative to psychological and normative realities. We must not expect too much precision of them. That is why premise 5 (§VI) requires only that the convict’s criminal history constitute prima facie evidence. However, premise 5 is true only if there is a positive, linear correlation between the number of prior convictions and culpability-aggravating character traits.

Some readers will read this condition and wonder how it could be doubted. Is it not obvious that repeat offenders are morally worse than first offenders, ceteris paribus? The first offender has made one blameworthy choice. The repeat offender has made more than one. The second is, to that extent, morally worse than the first offender. Compare two forty-year-olds, Laurel and Hardy, each of whom was recently convicted of burglary. The entirety of Laurel’s criminal record is fourteen prior convictions for burglary. Hardy has no prior convictions. The two have lived otherwise similar lives. Laurel chose to burgle fifteen times. Hardy chose to do so only once. Laurel made fifteen times as many blameworthy choices as Hardy. Making a blameworthy choice fifteen times is much worse (maybe about fifteen times worse) than making a similar choice only once. No retributivist denies that Laurel deserves more aggregate punishment over the course of their life than Hardy. Even flat-rate sentencing reflects this difference, giving the state fifteen sequential opportunities to punish Laurel for their fifteen choices, potentially putting Laurel behind bars for fifteen times as long as Hardy.Footnote ¹¹⁰

However, these facts do not support premise 5 of the CA unless they provide evidence of character differences. It is remarkably easy to slide from the obvious fact that Laurel’s record of choices is worse than Hardy’s to the much stronger conclusion that Laurel’s character is worse. Of course, one can restate the comparison in characterological language: Laurel has more of a ‘burgling character’ and is ‘more of a burglar’ than Hardy. Those statements are true if they mean simply that Laurel burgled more than Hardy. But culpability-relevant traits are not reducible to behavior (§VII.A). Rather, they are dispositional.Footnote ¹¹¹

Does Laurel have stronger burgling dispositions than Hardy? Yes, in a sense. For any convict, it is always possible to reference ‘whatever internal states functioned as but-for causes of his criminal choices.’ One need not know anything about these states to affirm their existence. If these states are relatively stable, relatively strong, and relatively consistent, they might constitute traits on my definition. Laurel’s record warrants inferring a ‘burgling trait.’ Hardy’s record warrants no such inference.

But the trait properly ascribed to Laurel, not Hardy, is not a candidate for culpability-aggravation unless it is also of a type that is relatively well-manifested in blameworthy conduct. And the crimes for which the defendant has already been punished do not count for this purpose, on pain of circularity. Laurel’s fourteen prior burglaries cannot count simultaneously as both 1) evidence that Laurel possesses a trait that is a cause of the fifteenth burglary; and 2) evidence that this trait aggravates culpability beyond the culpability for the burglaries themselves. Ascribing a culpability-aggravating trait on the basis of a criminal record is warranted only if the record provides reliable (if indirect) evidence of internal dispositions that are wider than the narrowest disposition compatible with the criminal record itself. Attributing a convict’s history to a particular negative character trait requires fitting their crimes into a larger conduct pattern that has predictive and/or explanatory power. Does Laurel have, for example, lower self-control (a TLSC) than Hardy? That is an open, empirical question.

Using revisionary traits further complicates the answer. A wider trait is, by nature, one for which a wider range of conduct constitutes evidence of possession. The narrower a trait, the less likely it is that a given conviction on the record constitutes evidence of its possession.

Here is another way to think about it. If we understand character traits as moderately strong (i.e., as dispositions to generate pro tanto inclinations), but non-trait factors predict criminal behavior more accurately than do such traits, then non-trait factors play a dominant role in determining the length and content of the criminal record itself. The record cannot then be used as reliable evidence of culpability-aggravating traits, so understood.

Stepping back, I can now articulate more completely what the CA needs from traits. It needs traits that are culpability-relevant, which entails both manifestation and agential implication. Manifestation, in turn, requires relative stability, relative width, and relative strength (implying criminogenesis). The traits must also be positively correlated with criminal history to a sufficient degree. Perfect bimodality or multi-modality is not required, but some clustering must emerge. The CA supports the FOD, for example, only if there is a sufficiently high correlation between second offenses and greater scalar culpability (holding constant severity of crimes). Going beyond the FOD to PLM, and then to the full-blown RP, we might also suppose that third offenders are more culpable than second offenders, and fourth more than third, ad infinitum. The CA needs traits that differentiate, not just offenders from non-offenders, but first offenders from second offenders and, possibly, tenth from eleventh.

I cannot disprove the existence of such traits. But the burden is on the CA proponents to describe them and produce evidence that recidivists possess them. Criminology might yet provide such evidence. A TLSC is one candidate. Other theories of criminality recognizing criminogenic personality traits should also be considered.

It would be informative to ascertain the degree of correlation, if any, between recidivism and other indicia of culpability-aggravating traits. How might we do so, as an empirical matter? Experimenters could examine, retrospectively, the life histories of a large sample of Florida Burglars (with comparable lifespans), distinguishing those who committed one burglary in their lifetimes from those who committed two. Blinded to the burglaries, themselves, they could ‘score’ the lifetime conduct records—including non-criminal conduct—based on observable evidence of culpability-relevant traits. (Again, a TLSC might or might not fit the bill.) Would they find statistically significant differences between the scores of the single-offender and double-offender populations? Between the double offenders and the triples? Does the fact that someone commits a second burglary provide evidence that they possess a larger number of culpability-aggravating traits than do first offenders, or that their traits are collectively more culpability-aggravating than those of first offenders?

Perhaps. I confess I find affirmative answers to such questions to be intuitively plausible. But I do not trust my intuitions in this area as much as I did before contemplating the character experiments. My intuitions also assured me that most of us have robust character traits that predict most of our behavior. My intuitions probably still reflect various folk-psychological assumptions: that recidivism manifests negative traits that are relatively wide, relatively strong, agency-implicating, and concentrated in the recidivist population. The experiments call into broad question the reliability of my intuitions about such matters. If the proponent of the CA invites me to ‘trust my gut’ on this, I must decline the invitation and seek empirical evidence. My preliminary examination of a TLSC has not yet persuaded me that there are empirically validated traits with the features required if the CA is to support even the FOD, much less PLM or the RP. The absence of such evidence is significant because many retributivists support the FOD and some support modest forms of PLM.

Such evidence might yet appear. But if it does, a countervailing concern arises. Given the right evidence, the CA could support the RP, perhaps even extreme versions. This would be a significant result because most retributivists oppose the RP, especially extreme versions. One message of this article is that anyone using the CA who favors modest forms of HSS (e.g., the typical FOD) over both flat-rate sentencing and more severe forms of HSS (e.g., the typical RP) must hope that just the right empirical evidence emerges.

This observation complements practical concerns others have expressed about proportionality in application. On what basis should the magnitude of the RP be determined? How might we set a non-arbitrary upper bound?Footnote ¹¹² I think it premature to try to answer such questions before we have the empirical facts needed to support HSS, simpliciter.

A broader objection to the CA would run as follows. Assume, for the moment, that we have isolated a particular character trait and concluded that it is culpability-aggravating. Assume, also, that there are statistically significant correlations between this trait and criminal records. Could we justify sentencing practices on the basis of mere correlations?Footnote ¹¹³

This objection raises a basic question about the implementation of retributive sentencing principles. If lawmakers wish to justify, in retributive terms, a sentencing rule that treats a particular attribute as a factor, how confident must they be that it constitutes reliable evidence of negative desert? There is no generally accepted answer to this question of non-ideal retributive theory.Footnote ¹¹⁴ If the standard is reasonable doubt, for example, experimental findings are unlikely to vindicate the CA. For any such trait, I do not anticipate experiments establishing that a repeat offender’s conviction record proves, beyond a reasonable doubt, that they possess it.

However, a reasonable doubt standard is probably too high for any retributivist who wants to endorse the use of sentencing factors in the real world. In any case, I have not addressed questions about how, if at all, sentencing guidelines should incorporate prior convictions. The conclusion of the CA is merely that prior convictions are prima facie evidence that is relevant to scalar culpability. If experimental findings persuade us to endorse the a posteriori premises of the CA, the question remains how much weight prior convictions should receive in guidelines. I would advise that the retributivist’s answer should reflect the strength of the experimental findings. The stronger the statistical correlation between prior convictions and culpability-aggravating traits, the greater the weight retributivists could justifiably direct courts to place on prior convictions, as a class. If the correlation is weak or non-existent, perhaps HSS is simply indefensible. If the correlation is modest, perhaps a weak and rebuttable presumption of aggravated culpability would be appropriate, subject to mitigating evidence concerning individual character and other culpability-relevant factors. Sentencing is an individualized process, by design, in which courts balance competing considerations.

XIV. Aggravated Culpability Without Bad Character

At this point, retributivists might reply that they never much liked character arguments, anyway. In this section, I consider arguments for HRD that replace premises 3-5 of the CA with premises that are ostensibly character-independent. The literature contains several such arguments, all of which face objections on their own merits. I shall contend that some of these arguments implicitly presuppose that criminal history correlates with specific character or personality traits. If so, these arguments may be vulnerable to the a posteriori objections of §§X-XIII, in addition to objections advanced by others.

XV. Implications and Directions for Further Inquiry

Retributivists who support HSS have generally rejected character-based arguments for philosophical reasons. I have tried to formulate such an argument (the CA) clearly and charitably. I conclude, perhaps surprisingly, that the CA is a valid argument for HSS in ideal theory. I found no compelling objections to its inferential validity or a priori premises. If retributivism is a viable, prescriptive theory of sentencing, then the CA could, in theory, support certain forms of HSS, maybe even some of the severer features found in the USA, which most scholars condemn.

In practice, however, the CA is sound only if two conditions hold: 1) convicts have culpability-aggravating traits; and 2) criminal history correlates properly with such traits and provides prima facie evidence thereof. Little work has been done to support the empirical hypotheses embedded in these premises. The character experiments challenge the prevalence of folk-psychological character traits, although the implications of this research are far from settled.

Although the traits needed by the CA need not be identical to the robust traits of folk psychology, I have suggested that they cannot be too revisionary if they are to be culpability-aggravating. Also, there is little evidence that criminal history is trait-correlated as the CA requires. For these reasons, I am unconvinced that either of the aforementioned conditions holds. Therefore, I have no confidence that the CA supports any version of HSS in the real world.

For similar reasons, I question any intuitive sense we have that repeat offenders have distinctive traits in virtue of which they deserve harsher penalties. Most retributivists grant considerable epistemic authority to our intuitive judgments about relative desert and associated ordinal and cardinal judgments about deserved sentences. However, our moral intuitions about these matters track our folk-psychological character judgments and likely reflect whatever inaccuracies afflict the latter. Voters, public officials, and theorists may have assumed that recidivism provides evidence of traits that are relatively wide and relatively strong—wider and stronger, perhaps, than the evidence supports. They may, therefore, have given too much weight to repeated offenses as culpability-aggravators.

Absent contrary evidence, perhaps it was reasonable to allow folk psychology to inform sentencing policy.Footnote ¹²¹ In the face of contrary evidence, however, the burden shifts to defenders of HSS and the penal status quo. If lawmakers consult public opinion as a guide to desert, it becomes more important than ever to correct public opinion if it proceeds from misconceptions about moral psychology.

One approach to reinforcing the CA would combine an empirically validated conception of culpability-relevant traits with an empirically validated model of the relationship between criminal record and traits, so conceived. Revisionist conceptions of personality traits exist. Criminology supplies some candidates for the job that aspire to empirical adequacy over fidelity to folk psychology, a TLSC being one example. The essential question is whether such traits, revisionary or not, are sufficiently relevant to culpability and correlated with criminal record to support any version of HSS via the CA. Criminogenesis is only one of the necessary attributes. The agential implication of a TLSC, in particular, has not been established.

Even if we conclude that the CA is sound, we must answer empirical questions to determine which versions of HSS it can support. If repeat offenders do not deserve longer sentences than first offenders, retributivists can endorse either flat-rate sentencing or, at most, the weakest forms of HSS—those with caps or plateaus no higher than what first offenders deserve. These are both politically awkward options because stronger forms of HSS remain popular with public officials and voters. Yet no one claims that first offenders actually deserve sentences as long as the caps found in many jurisdictions.

Someone committed to stronger forms of HSS might, instead, reject retributivism as a prescriptive sentencing principle. They might still be able to defend stronger HSS without claiming that repeat offenders actually deserve longer terms. Non-retributivists can try to defend it for crime-control reasons. Assuming repeat offenders are more likely to reoffend, longer periods of incapacitation might be justified for such reasons, even if such sentences exceed desert. Of course, there might not be good crime-control rationales for HSS, either, unless the efficacy of longer sentences is empirically validated.Footnote ¹²²

Anyone committed to both retributivism and stronger forms of recidivist sentencing must either defend a character argument or develop an alternative argument for the desert-relevance of criminal history, one that does not rely, either directly or indirectly, on traits as a negative desert base. I have raised concerns about the a posteriori premises of the character argument. Whether these concerns can be answered remains an open question.