PULSIFER v. UNITED STATES

SUPREME COURT OF THE UNITED STATES

_________________

No. 22–340

_________________

MARK E. PULSIFER, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the eighth circuit

---

---

 Justice Gorsuch, with whom Justice Sotomayor and Justice Jackson join, dissenting.

 The First Step Act of 2018 may be “ ‘the most significant criminal justice reform bill in a generation.’ ” Brief for Sen. Richard J. Durbin et al. as Amici Curiae in Terry v. United States, O. T. 2020, No. 20–5904, p. 9. Through the 1980s and 1990s, Congress adopted an ever-increasing number of ever-longer mandatory minimum prison sentences. In part due to these policies, the federal prison population grew by more than 100% in less than a decade. In the First Step Act, Congress sought to recalibrate its approach. It did so by promising more individuals the chance to avoid one-size-fits-all mandatory minimums and receive instead sentences that account for their particular circumstances and crimes.

 This dispute concerns who is eligible for individualized sentencing and who remains subject to mandatory minimums after the First Step Act. Before the Act, a defendant seeking to avoid a mandatory minimum had to satisfy five stringent statutory tests. After the Act, all those tests remain, only the first is now less demanding. As revised, it provides that a defendant may be eligible for individualized sentencing if he “does not have” three traits: (A) more than 4 criminal history points, (B) a 3-point offense, and (C) a 2-point violent offense. In lower court proceedings, the government admitted that this new test is “most natural[ly]” read to mean what it says: A defendant may be eligible for  individualized sentencing unless he possesses all three listed traits—A, B, and C. Brief for United States in  No. 19–50305 (CA9), p. 7 (Government CA9 Brief ); id., at 10–11; accord, Brief for United States in No. 21–1609 (CA8), p. 11 (Government CA8 Brief ). Despite its admission, however, the government urges us to adopt a different construction. It asks us to read the First Step Act as promising a defendant a chance at individualized sentencing only when he does not have any of the three listed traits—A, B, or C.

 If this difference seems a small one, it is anything but. Adopting the government’s preferred interpretation guarantees that thousands more people in the federal criminal justice system will be denied a chance—just a chance—at an individualized sentence. For them, the First Step Act offers no hope. Nor, it seems, is there any rule of statutory interpretation the government won’t set aside to reach that result. Ordinary meaning is its first victim. Contextual clues follow. Our traditional practice of construing penal laws strictly falls by the wayside too. Replacing all that are policy concerns we have no business considering. Respectfully, I would not indulge any of these moves.

I

A

 In approaching the dispute before us, some background helps. Before the 1980s, federal judges generally enjoyed broad discretion at sentencing. Often, they could impose punishments ranging from probation up to statutorily specified maximum prison terms. Mistretta v. United States, 488 U. S. 361, 363 (1989). In exercising that discretion, judges had to “consider every convicted person as an individual” and pick punishments that “fit the offender and not merely the crime.” Pepper v. United States, 562 U. S. 476, 487–488 (2011).

  Today, many defendants still receive individualized sentences. In the mine run of federal cases, a court will start with sentencing guidelines the United States Sentencing Commission has prepared at Congress’s direction. The guidelines help a court identify a range of presumptively reasonable sentences tailored to the defendant and his crime. See Rita v. United States, 551 U. S. 338, 347 (2007). That range depends on an “offense level,” a figure that takes into account the seriousness of the defendant’s crime and his role in it, as well as the defendant’s “criminal history” score, a tallying that accounts for his past misconduct. United States Sentencing Commission, Guidelines Manual §§1B1.1, 4A1.1–4A1.2, ch. 5, pt. A (Nov. 2023) (USSG); see Rosales-Mireles v. United States, 585 U. S. 129, 133–134 (2018). The guidelines, however, are just that. A sentencing judge may sometimes depart or vary from the guidelines’ recommended range, picking a lower or higher sentence if it best fits the defendant and broader penological goals Congress has instructed courts to consider. See Gall v. United States, 552 U. S. 38, 46, 49–50 (2007); 18 U. S. C. §3553(a).

 In the 1980s and 1990s, Congress pursued a different approach for certain drug offenses. See Anti-Drug Abuse Act of 1986, 100 Stat. 3207–2 to 3207–4; Anti-Drug Abuse Act of 1988, 102 Stat. 4370, 4377–4378. It required courts to impose mandatory minimum prison terms based only on the kind and quantity of the drugs involved in the defendant’s crime. A court “was required to send the offender to prison” for a set period of years “no matter how minor the offender’s participation in the offense may have been, and no matter what mitigating circumstances might have been present.” J. Rakoff, Why the Innocent Plead Guilty and the Guilty Go Free 13 (2021). Under this regime, for example, a defendant distributing 5 grams of crack cocaine faced a 5-year mandatory prison term, and one with 50 grams faced a 10-year term. 100 Stat. 3207–2 to 3207–3. Meanwhile, a  defendant found with powder cocaine confronted those same prison terms only if he distributed 100 times those amounts. Ibid.

 In short order, the federal prison population exploded. In 1986, federal prisoners numbered 30,104, approximately 37.7% of whom were serving time for drug offenses. Dept. of Justice, Sourcebook of Criminal Justice Statistics 519 (31st ed. 2003). By 1994, the federal prison population reached almost 74,000, with approximately 61.3% of inmates serving time for a drug offense. Ibid.

 Calls for reform came quickly and grew with time. See, e.g., U. S. Sentencing Commission, Special Report to the Congress: Mandatory Minimum Penalties in the Federal Criminal Justice System iii (1991); id., App. G (collecting statements from the Judicial Conference and 12 circuits). Eventually, Congress responded to these calls in various ways. In one reform, for example, it prospectively reduced the crack-cocaine disparity from 100:1 to 18:1. See Fair Sentencing Act of 2010, 124 Stat. 2372. In another, it adopted §3553(f ), a provision that came to be called the “safety valve” and that lies at the heart of today’s case. See Violent Crime Control and Law Enforcement Act of 1994, 108 Stat. 1985–1986.

 As originally enacted in 1994, the safety valve provided modest relief. It exempted defendants who could meet five statutory criteria from otherwise-applicable mandatory minimums, directing instead that they should receive individualized sentences. Ibid. (codified as amended at 18 U. S. C. §3553(f )). But the first of the safety valve’s five criteria, codified in paragraph (f )(1), was especially demanding. It precluded relief for any individual with “more than 1 criminal history point”—meaning that a defendant could find himself ineligible for individualized sentencing if his background included even a single 60-day prison term or two prior offenses involving no prison term at all. 108 Stat. 1985; see §3553(f )(1) (1994 ed.); USSG §§4A1.1(b)–(c),  4A1.2 (Nov. 1994).

B

 In the First Step Act of 2018, Congress adopted an array of further reforms. Pub. L. 115–391, 132 Stat. 5194. Passed with overwhelming majorities in both chambers of Congress and with presidential support, the Act reduced the length of some mandatory minimums by 25 percent. See §401, id., at 5220–5221. It narrowed the circumstances under which a court could “stack” certain mandatory minimums on top of one another. See §403(a), id., at 5221–5222; U. S. Sentencing Commission, The First Step Act of 2018: One Year of Implementation 5 (2020). And it made Congress’s earlier amendment to the crack-cocaine disparity retroactive, allowing individuals sentenced before that amendment’s adoption a chance at resentencing. See §404, 132 Stat. 5222.

 The First Step Act also revised the safety valve’s first provision. Where paragraph (f )(1) once barred a defendant with even a single criminal history point from receiving an individualized sentence, Congress now chose a different course. As amended, the full safety valve today instructs a court to afford an individualized sentence “if [it] finds at sentencing . . . that—”

 “(1) the defendant does not have—

 “(A) more than 4 criminal history points, excluding any criminal history points resulting from a 1-point offense, as determined under the sentencing guidelines;

 “(B) a prior 3-point offense, as determined under the sentencing guidelines; and

 “(C) a prior 2-point violent offense, as determined under the sentencing guidelines;

 “(2) the defendant did not use violence or credible threats of violence or possess a firearm or other dangerous weapon (or induce another participant to do so) in connection with the offense;

  “(3) the offense did not result in death or serious bodily injury to any person;

 “(4) the defendant was not an organizer, leader, manager, or supervisor of others in the offense, as determined under the sentencing guidelines and was not engaged in a continuing criminal enterprise, as defined in . . . the Controlled Substances Act; and

 “(5) not later than the time of the sentencing hearing, the defendant has truthfully provided to the Government all information and evidence the defendant has concerning the offense or offenses that were part of the same . . . common scheme or plan . . . .” 18 U. S. C. §3553(f ).

C

 The question we face concerns how the amended safety valve works. Everyone agrees that a defendant must still clear five daunting statutory hurdles. But the parties disagree what the first entails after the First Step Act. Observing that the word “and” connects each of the subparagraphs (f )(1)(A), (B), and (C), Mark Pulsifer argues that the safety valve’s first provision now operates to render ineligible one kind of defendant—a defendant who bears all three enumerated traits, A, B, and C. Because he does not have all three, Mr. Pulsifer submits, he is eligible for safety-valve relief as long as he can satisfy the law’s four remaining provisions. Meanwhile, on the government’s telling, paragraph (f )(1) renders three kinds of defendants ineligible for relief—any defendant who has trait A, B, or C. And because Mr. Pulsifer has at least one of those traits, the rest of the safety valve is irrelevant; paragraph (f )(1) alone renders him ineligible for relief.

 Disputes about the amended safety valve’s operation have simmered for years in the lower courts and yielded  conflicting results.¹ At least one thing, though, is clear: The dispute before us matters profoundly. According to a Sentencing Commission analysis based on 2021 data, about 33% of drug offenders were eligible for safety-valve relief under the law’s old terms. See 88 Fed. Reg. 7186 (2023). Under Mr. Pulsifer’s understanding of the First Step Act, about 66% would become eligible for individualized sentencing. See ibid. By contrast, under the government’s reading of the Act, that number would shrink to around 44%. See ibid. Our decision today thus promises to affect the lives and liberty of thousands of individuals.

II

 Unless some feature of the law suggests that one or another of its terms bears a specialized meaning, our duty is to interpret Congress’s work as an ordinary reader would. See Niz-Chavez v. Garland, 593 U. S. 155, 163 (2021). At the heart of today’s dispute lies no specialized term but perhaps the most ordinary of words: Everything turns on what work the word “and” performs in paragraph (f )(1), where a sentencing court is tasked with determining whether “the defendant does not have” three traits—A, B, “and” C.

A

 In taking up the parties’ dispute, start with a few simple and uncontested observations. First, as the Court agrees, “and” is “a conjunction—a word whose function is to connect specified items.” Ante, at 7; see J. Opdycke, Harper’s English Grammar 200 (rev. ed. 1966).

  Second, and more specifically, “and” is an “additive” conjunction, one often indicating that the words it connects should be added together. Id., at 200; The Chicago Manual of Style §5.183, p. 191 (15th ed. 2003). As the Court explains, when “and” performs that role, it means “[t]ogether with,” “along with,” “in addition to,” or “as well as.” American Heritage Dictionary 66 (5th ed. 2018); see ante, at 7.

 Third, in paragraph (f )(1) “and” connects a list in a negative conditional statement (“if . . . the defendant does not have”). Negative conditional “if . . . not” statements often function like the word “unless.” See R. Huddleston & G. Pullum, The Cambridge Grammar of the English Language §14.3, p. 755 (2002). Consider the mother who tells her child, “If you do not have any homework left, you can go play with your friends.” The child would understand that he could play with his friends unless he had homework left to do.

 Now apply those observations to paragraph (f )(1). Given the meaning of “and,” an ordinary reader would naturally understand that a defendant is eligible for individualized sentencing if he “does not have” trait A, trait B, together with trait C. Add to the mix what we know about the interchangeability of “if . . . not” and “unless”: A defendant may receive guidelines sentencing unless he has trait A, trait B, together with trait C. Put the points together, and the statute indicates that a court may issue an individualized sentence unless the defendant has all three traits listed in §3553(f )(1), just as Mr. Pulsifer contends.

B

 What the language of paragraph (f )(1) suggests, surrounding context confirms. When Congress uses different terms in a statute, we normally presume it does so to convey different meanings. Southwest Airlines Co. v. Saxon, 596 U. S. 450, 457–458 (2022). We sometimes call this presumption the “meaningful-variation canon.” Id., at 457.  Here, we see just such a meaningful variation. When Congress sought a single word to indicate that one trait among many is sufficient to disqualify an individual from safety-valve relief, it chose an obvious solution: not the conjunctive “and,” but the disjunctive “or.”

 In fact, Congress used “or” this way no fewer than three times. Paragraph (f )(2) specifies that, for a defendant to be eligible for individualized sentencing, a court must find that “the defendant did not use violence or credible threats of violence or possess a firearm or other dangerous weapon (or induce another participant to do so) in connection with the offense.” (Emphases added.) Paragraph (f )(3) premises eligibility on a finding that a defendant’s “offense did not result in death or serious bodily injury to any person.” (Emphasis added.) And paragraph (f )(4) provides that eligibility for relief turns on whether the defendant “was not an organizer, leader, manager, or supervisor of others in the offense.” (Emphasis added.)

 The fact that Congress repeatedly used “or” when it wanted relief to turn on a single trait among many suggests that the “and” in paragraph (f )(1) performs different work. Even the government once acknowledged as much, conceding below that the “and” in paragraph (f )(1) is “most natural[ly]” read as requiring a sentencing court to find that a defendant possesses all three listed traits before holding him ineligible for relief. Government CA9 Brief 7; id., at 10–11; accord, e.g., Government CA8 Brief 11. Nor is the government alone in this unsurprising understanding: A study involving ordinary Americans found that the largest share of participants understood a sentence tracking paragraph (f )(1)’s structure to trigger ineligibility only if all three conditions are satisfied. See Brief for Thomas R. Lee et al. as Amici Curiae 15, 18.

 III

A

 The government disputes none of this evidence about the law’s ordinary meaning. Instead, it begins with a theory. Maybe, the government says, there is another “permissible” way to read paragraph (f )(1). Ante, at 7; Brief for United States 18, 37. Maybe Congress implicitly wanted a reader to “distribut[e]” the “verb phrase” “does not have” among each subparagraph. Ante, at 8–9 (internal quotation marks omitted); Brief for United States 14–18. Maybe, then, we should effectively read the statute to work this way, with deleted words stricken and new ones added in bold:

(1) the defendant does not have —

 (A) does not have more than 1 criminal history point . . . ;

 (B) does not have a prior 3-point offense . . . ; and

 (C) does not have a prior 2-point violent offense.

Yes, the government’s implicit distribution theory requires a reader to delete words before the em dash. Yes, it requires a reader to reinsert them in three different places where they do not appear. But maybe, the government suggests, Congress implicitly intended for a reader to do all that. Even though what it wrote is susceptible to a far more natural construction requiring none of these gymnastics.

 That is not how statutory interpretation usually works. Statutes aren’t games or puzzles but “instruments of a practical nature, founded on the common business of human life, . . . and fitted for common understandings.” 1 J. Story, Commentaries on the Constitution of the United States §451, p. 437 (1833). For that reason, we usually presume that Congress “employed words in their natural sense, and . . . intended what [it] said.” Gibbons v. Ogden, 9 Wheat. 1, 188 (1824). And once we have identified the most natural sense of the law’s terms, as we have here, our interpretive task is usually at an end. See, e.g., Barnhart v. Sigmon Coal Co., 534 U. S. 438, 461–462 (2002).

 The government’s implicit distribution theory is so far from the most natural reading of the law that its many and able lawyers didn’t even stumble on it until late in the game. In litigation below, the government started by arguing primarily that paragraph (f )(1) “must be read in the disjunctive”—a fancy way of saying that “and” means “or.” Government CA8 Brief 4; see United States v. Garcon, 54 F. 4th 1274, 1280 (CA11 2022) (en banc). In early cases, that was the government’s only argument. See, e.g., Sentencing Tr. in No. 3:19–cr–207 (ED Tenn.), ECF Doc. 176, p. 4 (“I think the Department of Justice’s position as well as our position here today is . . . that it should be read disjunctively”); see also Tr. of Oral Arg. 103. Only after a resounding loss on that argument, see United States v. Lopez, 998 F. 3d 431, 435–443 (CA9 2021), did the government shift to its implicit distribution theory, stressing that its new offering does not require courts to “transform” “and” into “or,” see Brief for United States 42–43.

 The government’s implicit distribution theory may be a “convenient litigating position,” Bowen v. Georgetown Univ. Hospital, 488 U. S. 204, 213 (1988), but it does not come close to respecting the most natural construction of the law. It may have the benefit of leaving “and” alone, but it comes at the cost of rearranging so much else in the statute. One way or another, the government cannot get where it wishes to go without tinkering with the law. And to know that much should be enough to bring this case to a close: “Crimes are supposed to be defined by the legislature, not by clever prosecutors riffing on equivocal language.” Dubin v. United States, 599 U. S. 110, 129–130 (2023) (internal quotation marks and alteration omitted).

B

 How does the government reply? It insists that contextual clues support its implicit distribution theory. These  clues are so compelling, it says, any other construction of the law isn’t “plausible” or “possible.” Ante, at 16, 27; Brief for United States 18–19. It is a bold claim, not only because the government overlooks all the evidence of the statute’s meaning outlined above, but also because it overlooks one piece of contextual evidence after another weighing against its implicit distribution theory.

 Start with this one: The statute before us stands far afield from classic cases that invite questions about implied distribution. In everyday speech, the government stresses, a listener may appreciate the need to “distribut[e]” what this Court has called “several antecedents” to “several consequents.” Encino Motorcars, LLC v. Navarro, 584 U. S. 79, 87 (2018) (quoting 2A N. Singer & S. Singer, Sutherland Statutes and Statutory Construction §47:26, p. 448 (rev. 7th ed. 2014) (internal quotation marks omitted)). In its brief before us, the government leads with this example of the phenomenon: someone who says, “I sell red, white, and blue caps.” See Brief for United States 14. That statement, the government observes, contains an ambiguity. One listener might think that the seller offers caps bearing all three colors. But another listener might wonder if the seller implicitly means to “distribute” different colors to different caps—so that she really means to say she sells red caps, she sells blue caps, and she sells white caps. Only context, the government insists, can resolve the ambiguity and reveal which understanding best reflects the seller’s meaning. Id., at 16.

 If context suggests anything, however, it is that this observation has little to offer when it comes to the statute before us. The First Step Act does not contain several “antecedents” (many caps, for example) that might or might not distribute among several “consequents” (say, colors). Instead, paragraph (f )(1) speaks of a single person—“the defendant” presently before the sentencing court—who must  not have three specified traits (A, B, and C). And that “singular” construction “tends to avoid the ambiguity” about distribution that a “plural” construction can invite. M. Kirk, Legal Drafting: The Ambiguity of “And” and “Or,” 2 Tex. Tech. L. Rev. 235, 239–240 (1971); see also Huddleston, Cambridge Grammar §1.3.1, at 1280–1281.

 Drafting experts illustrate the point with this phrase: “charitable and educational institutions.” R. Dickerson, The Fundamentals of Legal Drafting §6.2, pp. 109–110 (2d ed. 1986); Kirk, 2 Tex. Tech. L. Rev., at 239–241. The phrase is ambiguous. The multiple “institutions” might distribute across the multiple listed traits to describe both “charitable institutions and educational institutions.” Dickerson, Fundamentals of Legal Drafting §6.2, at 110; Kirk, 2 Tex. Tech. L. Rev., at 240. Or the term “institutions” might not distribute, so the phrase describes only institutions that are both charitable and educational. Id., at 240–241. But if there is just a single “institution,” any ambiguity dissipates: “A charitable and educational institution” is an institution with both traits. The same holds true when a saleswoman offers “the red, white, and blue cap”: In that case, a buyer knows with certainty that the seller offers one kind of cap bearing all three colors.

 This contextual clue poses the government with a serious problem. When Congress wrote paragraph (f )(1), it employed a singular construction that tends to avoid the ambiguity about distribution that plural constructions invite. The statute before us thus bears no resemblance to the government’s lead illustration involving multiple caps and colors. Nor does it bear any resemblance to the government’s various illustrations from statutory and constitutional law involving multiple “offenses” that fall into multiple classes, see ante, at 11; Brief for United States 17–18 (discussing 34 U. S. C. §20101(f )); multiple “Cases” that meet multiple descriptions, see ante, at 8–9; Brief for United States 40 (quot ing Art. III, §2, cl. 1); or the many kinds of “Commerce” Congress can regulate, see ante, at 9; Brief for United States 39–40 (quoting Art. I, §8, cl. 3).²

 Sensing the government’s difficulty, the Court struggles for an example of its own involving a singular person or thing that does generate an ambiguity about distribution. Eventually, it lands on Eric Carle’s story about a caterpillar who “ ‘ate through’ ” (among so many other things) “ ‘one sausage, one cupcake, and one slice of watermelon.’ ” Ante, at 8; see also ante, at 9, n. 3. Mission accomplished: One child might implicitly distribute the phrase “ate through” to each foodstuff, while another might read the list without implicit distribution to mean the caterpillar ate through a “combination” that includes them all. Ante, at 9.

 But what does that prove? “[T]o acknowledge ambiguity is not to conclude that all interpretations are equally plausible.” Gwaltney of Smithfield, Ltd. v. Chesapeake Bay Foundation, Inc., 484 U. S. 49, 57 (1987). And an example of ambiguity about distribution in a children’s book does nothing to prove that the federal criminal statute before us is most plausibly read to require implicit distribution. Add some of paragraph (f )(1)’s salient features into the illustration and that much becomes clear. As the story goes, the caterpillar is in the process of becoming a butterfly. So suppose the story said the caterpillar “will remain a caterpillar if he does not eat (A) one sausage, (B) one cupcake, and (C) one slice of watermelon.” I suspect most ordinary readers (and children) would have little trouble concluding that the sentence means that the caterpillar will remain a caterpillar unless he eats all three things; one alone will not do.

 C

 Here’s another problem with the government’s theory: If in some contexts a speaker might intend for a listener to distribute words implicitly, the context before us counsels against attributing any such intention to Congress. It does because a careful look at the safety-valve statute reveals that, when Congress wanted to distribute a phrase in this law, it did not leave the matter to implication. It did not depend on the reader’s imagination. Instead, Congress distributed phrases expressly.

 Twice, in fact. In paragraph (f )(4), Congress took the trouble to distribute expressly the phrase “was not,” permitting relief only if “the defendant was not an organizer, leader, manager, or supervisor of others . . . and was not engaged in a continuing criminal enterprise.” (Emphases added.) Likewise, in paragraph (f )(1) itself Congress expressly distributed the phrase “as determined under the sentencing guidelines” three times, in each of subparagraphs (A), (B), and (C). All the contextual evidence before us thus suggests that, in a statute carrying grave criminal consequences, Congress was careful with its words and concerned with clarity. It did not leave ambiguities about distribution to be resolved by implication. Instead, it resolved them expressly, even at the cost of repetition.

 Once more, the government’s examples only serve to illustrate its problem. It imagines a speaker who says, “ ‘I didn’t like his mother and father.’ ” Ante, at 10; Brief for United States 39. The government suggests that a listener would “probably” understand the sentence as implicitly distributing the phrase “I didn’t like his,” so that it really means, “I didn’t like his mother and I didn’t like his father.” Ante, at 10 (emphasis added); Brief for United States 39. But as the hedge (“probably”) indicates, an ambiguity lurks here. The sentence could also be understood without any distribution to convey the idea that “I didn’t like his mother and father” as a couple, even if I liked each individually well  enough. See Huddleston, Cambridge Grammar §2.2.2, at 1298–1299.³ Only context, the government concedes, can clarify which meaning is more apt. See ante, at 15; Brief for United States 16. Yet somehow, the government neglects that same message when it comes to the statute before us—where context reveals that Congress did not leave questions of distribution to implication but resolved them expressly.

D

 Context exposes yet another flaw in the government’s implicit distribution theory. If, as the government imagines, Congress was determined to find an “efficient” way to disqualify a defendant bearing any one of the three traits listed in paragraph (f )(1), ante, at 8, 13; see Brief for United States 18, it had an obvious solution before it: the word “or.” As we have seen, Congress employed exactly that approach three times in the safety valve: Paragraphs (f )(2), (f )(3), and (f )(4) all premise disqualification for relief on the presence of one trait or another. See Part II–B, supra. In this way, too, context confirms that, when Congress wanted to make one trait among many disqualifying, it proceeded expressly (and often efficiently)—but never by implication.

 After disregarding others, the government at least acknowledges this particular complication for its theory. It responds this way: Even substituting “or” for “and,” it says, would not “delive[r] us from interpretive controversy.” Ante, at 12; Brief for United States 26. It would not because replacing “and” with “or” in paragraph (f )(1) still would not  answer the question whether a single trait alone is enough to render a defendant ineligible for relief. Ante, at 12; Brief for United States 26. As evidence of the malleability of the word “or” in some contexts, the Court cooks up various illustrations involving a hypothetical chef. Ante, at 12, n. 5.

 It is a remarkable response. At argument, the government acknowledged that “or” “might have been a clearer way to express” that a single trait is disqualifying in paragraph (f )(1). Tr. of Oral Arg. 98. Below, the government initially pushed for treating “and” as meaning “or” precisely because it knew that doing so would mean that a defendant is ineligible for relief if he has even one of its listed traits. See id., at 101; Government CA9 Brief 11–13; Government CA8 Brief 7–8. And everyone, the Court included, concedes that Congress’s use of the word “or” in paragraph (f )(4) means that a defendant meeting any one of several criteria is disqualified from relief. Ante, at 25. Simply put, “we wouldn’t be sitting here if Congress had used the word ‘or’ ” in paragraph (f )(1). Tr. of Oral Arg. 97. Whatever ambiguity “or” might carry in other contexts, it carries none in §3553(f ). Throughout the safety valve, Congress used it to indicate that a single trait among many is disqualifying.⁴

  Finding the government on its back foot, the Court again comes to its defense, this time by trying to change the rules of play. Perhaps, the Court speculates, Congress’s choice of “and” rather than “or” in paragraph (f )(1) was the product of careless drafting. See ante, at 11–12. Perhaps, too, those two conjunctions are “versatile” words not entitled to the respect we usually pay Congress’s variations in usage—a respect, the Court suggests, that is due only “to terms with some heft and distinctiveness, whose use drafters are likely to keep track of and standardize.” Ante, at 23, 25.

 Consider how far we have now retreated. Lower courts rejected the government’s and-means-or argument. In response, the government introduced its implicit distribution theory. Before us, the government stresses that its new theory does not depend on “transform[ing]” “and” into “or.” Brief for United States 42; see also id., at 15, 25. At first, the Court seems to proceed on the same premise. See ante, at 7–8. But now it reverses course. Resuscitating an argument the government itself has abandoned, the Court contends not just that the terms “and” and “or” are interchangeable, but that we need not even rely on our usual rules of interpretation when faced with them.

 This argument was a loser below and it should be here. When Congress employs “differing language in . . . two subsections,” we start from a presumption that it meant to convey a difference in meaning, not a presumption that it made “a simple mistake in draftsmanship.” Russello v. United States, 464 U. S. 16, 23 (1983). Never, to my knowledge, has this Court suggested that we may turn our back on this approach when conjunctions or other putatively “indistinctive” words are in play. Nor have we deployed that approach for “hefty” words alone—as if we were picking paper towels instead of interpreting statutes. To the contrary, our cases begin (and often end) with the presumption that Congress is careful in all its word choices and afford variations  between terms like “and” and “or” the same respect due others. See, e.g., United States v. Palomar-Santiago, 593 U. S. 321, 326 (2021) (reversing the lower court for failing to give effect to a statute’s use of “the conjunctive ‘and’ ”); Encino Motorcars, 584 U. S., at 87 (resting a reading of the relevant statute on “the ordinary, disjunctive meaning of ‘or’ ”); Loughrin v. United States, 573 U. S. 351, 357 (2014) (rejecting an argument that would “disregard what ‘or’ customarily means”); Reiter v. Sonotone Corp., 442 U. S. 330, 338–339 (1979) (similar); Rumsfeld v. Padilla, 542 U. S. 426, 434 (2004) (giving weight to the federal habeas statute’s “consistent use of the definite article”).⁵

 Nor could the premise latent in the Court’s argument be further from the truth. The difference between words like “and” and “or” often cannot be easily dismissed as meaningless when it comes to settling legal rights. Just imagine if the Sixth Amendment gave the accused a “right to a speedy or public trial.” Rather than getting a both timely and transparent trial, a defendant would be forced to choose which feature he prefers. Because the difference between “and” and “or” so regularly proves dispositive of important legal rights, drafting manuals for legal text from contracts to congressional legislation warn about the need to deploy the terms with care. See, e.g., Senate Office of the Legislative Counsel, Legislative Drafting Manual 64–65 (1997); K. Adams, A Manual of Style for Contract Drafting §§11.9–11.11, p. 211 (3d ed. 2013). And here, of course, the difference between “and” and “or” affects the lives of thousands, see supra, at 7—a fact so inconvenient for the Court that the Court says to ignore it as well, see ante, at 26.⁶

 IV

 So far, things look bleak for the government. Mr. Pulsifer offers a perfectly natural reading of the law. In response, the government offers a theory that it says rises or falls based on context. See ante, at 15; Brief for United States 11, 16. Yet, as it turns out, not one but three contextual clues array against its theory.

 Unable to muster a convincing response to any of that, the government pivots. Even if its implicit distribution theory suffers so many flaws, the government urges us to adopt it anyway because Mr. Pulsifer’s reading of the law would introduce a superfluity into the safety-valve statute. It is a resourceful reply. The government has many. But it, too, falls flat.

A

 Without question, the canon against superfluity can be a useful tool when seeking the meaning of a statute. It rests on the same principle as the canon of meaningful variation: the presumption that Congress is a careful drafter and each word it chooses “is there for a reason.” Advocate Health Care Network v. Stapleton, 581 U. S. 468, 477 (2017). But that fact also makes the government’s choice to rest its case on the superfluity canon a curious one. As we have seen,  the government’s implicit distribution theory depends on the assumption that Congress was not a careful drafter. It requires us to assume Congress left a distribution implicit in one section of paragraph (f )(1), even as it made others express elsewhere in paragraphs (f )(1) and (f )(4). It requires us to assume Congress meant for “and” in paragraph (f )(1) to do the same work as “or” in paragraphs (f )(2)–(f )(4). Sometimes, it seems, we are supposed to assume Congress was sloppy, other times careful. The only common thread seems to be what benefits the government in the moment.

 Even putting that small irony aside, the government has a bigger problem: Mr. Pulsifer’s reading leaves no provision in this statute superfluous. As the government sees it, a defendant who has both the prior 3-point offense required by subparagraph (B), and the 2-point violent offense required by subparagraph (C), will necessarily have more than the 4 criminal history points required by subparagraph (A). Because of this, the government submits, subparagraph (A) has no work to perform on Mr. Pulsifer’s reading: “Remove it from the statute, and what is left will make the exact same people eligible (and ineligible) for relief.” Ante, at 17; Brief for United States 19–20. Only its implicit distribution theory, the government contends, can cure the problem by allowing one subparagraph to “disqualif[y] defendants from relief even when the others would not.” Ante, at 16; Brief for United States 19–20.

 It’s a nice argument, but it rests on a faulty premise. As it happens, a defendant who has a 3-point offense under subparagraph (B) and a 2-point violent offense under subparagraph (C) often will not have “more than 4 criminal history points . . . under the sentencing guidelines” for purposes of subparagraph (A). And in cases like that, subparagraph (A) performs vital work under Mr. Pulsifer’s reading of the law by ensuring that the defendant remains eligible for relief. There is simply no surplus here for the  government’s implicit distribution theory to cure.

 To appreciate why this is so, consider the sentencing guidelines Congress cross-referenced in subparagraphs (A) through (C). They set forth a two-step process for calculating a defendant’s criminal history. At the first step, discussed in §4A1.1 of the guidelines, a judge assigns points to the defendant’s prior offenses. Usually, the points correspond to the length of the defendant’s previous sentences. So, for example, three points normally attach to an offense carrying a sentence longer than 13 months, two points to an offense with a sentence shorter than that but at least 60 days long, and one point to any other sentence.

 At the second step, described in §4A1.2 of the guidelines, a judge then computes the defendant’s criminal history. But during this process, a judge doesn’t just tote up all the points assigned to each offense. Under a variety of circumstances, the guidelines instruct a judge not to count points assigned to one offense or another. Points associated with hitchhiking, public intoxication, and fish and game offenses, for example, “are never counted.” §4A1.2(c)(2). Nor are points associated with sentences imposed by a court-martial, a foreign court, or a tribal court. §§4A1.2(g)–(i). The guidelines also instruct judges not to count points associated with offenses of a certain age. So, by way of illustration, if the defendant finished his sentence for a 3-point offense more than 15 years ago, those points are not counted. Likewise, if the defendant finished his sentence for a 2-point offense more than 10 years ago, those points do not count. §§4A1.2(e)(1)–(3). Courts thus perform “a single calculation” of a defendant’s criminal history score. Ante, at 19, n. 6. But in doing so, they routinely distinguish between the points an offense carries and a defendant’s ultimate, countable criminal history points.⁷

  Now return to subparagraph (A). It provides that the defendant must not have “more than 4 criminal history points, excluding any criminal history points resulting from a 1-point offense, as determined under the sentencing guidelines.” §3553(f )(1)(A) (emphasis added). As the italicized language demonstrates, when adopting the First Step Act Congress fully appreciated the distinction between what points an offense carries and whether those points contribute to a defendant’s criminal history score. And because of that very distinction, it is possible for a defendant to have a prior 3-point offense and a prior 2-point violent offense without having more than 4 criminal history points. Most obviously, as Chief Judge Pryor, former Acting Chair of the Sentencing Commission, has observed, a defendant may have a 3-point offense and a 2-point violent offense but both offenses are so old that he scores no criminal history points at all. See USSG §§4A1.2(e)(1)–(3); Garcon, 54 F. 4th, at 1281. As Judge Wood has noted, there are a variety of other situations as well in which a defendant will have both a 3-point offense and a 2-point violent offense but still not have more than four criminal history points. See United States v. Pace, 48 F. 4th 741, 763–764 (CA7 2022) (dissenting opinion).

 To know that is to know no superfluity problem exists—and thus no need to resort to the government’s implicit distribution theory to solve it. On Mr. Pulsifer’s reading of the  law, a court applying subparagraph (A) will consult the sentencing guidelines’ methodology for scoring criminal history points set forth in §4A1.2. In doing so, the court may find that, while the defendant has a prior 3-point offense and a prior 2-point violent offense for purposes of subparagraphs (B) and (C), one or another is too old or suffers from some other flaw so that he does not have more than four criminal history points. In all these cases, subparagraph (A) does significant work by making clear that, despite having a prior 3-point offense and a prior 2-point violent offense, the defendant remains eligible for relief. See Garcon, 54 F. 4th, at 1281–1282; Pace, 48 F. 4th, at 763 (opinion of Wood, J.).

B

 The government does not contest the central observation that defeats its superfluity argument. It admits that certain past offenses “ad[d] zero points to [a defendant’s] criminal-history score.” Brief for United States 32, n. 2; ante, at 17. So what exactly is the problem here?

 To complain about a superfluity problem, it turns out the government must create one. It does so this way. As written, subparagraphs (B) and (C) require a sentencing court to ask whether the defendant “ha[s]” a “3-point offense, as determined under the sentencing guidelines,” and “a 2-point violent offense, as determined under the sentencing guidelines.” But, the government suggests, we should read those provisions differently. We should read them to require a sentencing court to ask the further question whether the defendant’s offenses also score criminal history points. As the government candidly admits, its superfluity argument depends on reading subparagraphs (B) and (C) as “car[ing] only about offenses that do score . . . criminal- history points.” Brief for United States 28–29; ante, at 19–20. Only then might subparagraph (A) be left without work to perform, for indeed an offense that scores three criminal history points under subparagraph (B) and a violent offense  that scores two criminal history points under subparagraph (C) will always score more than four criminal history points under subparagraph (A).

 Put plainly, for the government’s superfluity argument to gain any traction, we must read still more words into the First Step Act, construing it now this way:

 (1) the defendant does not have —

 (A) does not have more than 4 criminal history points, excluding any criminal history points resulting from a 1-point offense, as determined under the sentencing guidelines;

 (B) does not have a prior 3-point offense that scores 3 criminal history points, as determined under the sentencing guidelines; and

 (C) does not have a prior 2-point violent offense that scores 2 criminal history points, as determined under the sentencing guidelines.

 It is one more remarkable request. Last I heard, the canon against assuming Congress has adopted superfluous words is not a license for judges to create a superfluity by inserting new words into a law. Let alone do so simply to help the government make its implicit distribution theory seem just a little less implausible.

V

 At this stage, the government withdraws to its final redoubt: a policy argument. In the government’s view, the only “function” Congress gave paragraph (f )(1) was the task of separating “more from less serious prior offenders.” Ante, at 21–22; Brief for United States 21. Affording the statute’s terms their ordinary meaning, the government asserts, would not allow the law to perform that “purpose” adequately. Brief for United States 20. By contrast, its implicit distribution theory would enable the law to fulfill its intended “role” “unerringly.” Ante, at 21, 22, n. 7; Brief for  United States 21.

 If this policy argument sounds familiar, it is because we have time and again rejected ones just like it. We do not presume that a law performs only one “function” or “role,” but recognize that almost every piece of legislation seeks to serve many competing purposes. See Luna Perez v. Sturgis Public Schools, 598 U. S. 142, 150 (2023); Barnhart, 534 U. S., at 461; Chicago v. Environmental Defense Fund, 511 U. S. 328, 339 (1994). We do not suppose that a law pursues any of those competing purposes to its logical end, acknowledging instead that almost every law is the product of compromise. Luna Perez, 598 U. S., at 150. And we do not displace ordinary statutory terms with judicial “speculation as to Congress[’s] intent,” Magwood v. Patterson, 561 U. S. 320, 334 (2010), because the American people have consented to be governed by the written laws their elected representatives adopt, not by the conjecture of others, see United States v. Bass, 404 U. S. 336, 348 (1971). For all these reasons and more, “it is quite mistaken to assume,” as the government does, “that whatever might appear to further the statute’s primary objective must be the law.” Henson v. Santander Consumer USA Inc., 582 U. S. 79, 89 (2017) (internal quotation marks and alterations omitted).

 Perhaps recalling our frequent admonition that policy talk cannot overcome plain text, the government tries a bit of rebranding. Although it refers occasionally to the First Step Act’s “purpose,” Brief for United States 20–21, 34, 48, for the most part it frames its argument in terms of rationality. When we measure the competing interpretations before us against how well they perform the statute’s only job, the government insists, we will find that the law’s ordinary meaning invites “arbitrar[y]” results and “nonsensical” implications. Id., at 22, 34, 36, 48. The Court buys into this thinly disguised policy appeal, see ante, at 20–22, and n. 7, even as it forcefully (and without a trace of irony) faults Mr. Pulsifer for appealing to statutory “purpose,” ante, at 26.  New framing or old, however, we have no business entertaining the government’s ramshackle argument.

 If anything, the government’s attempt at rebranding only makes matters worse for it. When a statute produces a truly irrational result, we have a doctrine to deal with the dilemma: absurdity. In narrow circumstances, a simple and “eas[ily]” fixed statutory error that “no reasonable person could intend” may be amenable to judicial correction under this Court’s traditional absurdity doctrine. See A. Scalia & B. Garner, Reading Law: The Interpretation of Legal Texts 234, 237 (2012); Story, Commentaries §427, at 411. It is a highly demanding doctrine—deliberately so, for judges have no license to rewrite a law’s terms just because they happen to think different ones more sensible. And, tellingly, no one thinks this law produces anything like an absurd result that might call for a judicial remedy. In fact, the government affirmatively disavows any reliance on absurdity doctrine. See Brief for United States 36. Instead, it only gestures vaguely in the direction of “nonsensical” results and asks us to run with the idea. As if we could tinker with Congress’s work on the basis of some newly fashioned  “absurdity-lite” doctrine.

 There is a reason why the government does not attempt an argument actually grounded on absurdity doctrine. Its core complaint is that the natural reading of the law does not, with sufficient precision, separate “more from less serious prior offenders.” Ante, at 22; Brief for United States 21. But, of necessity, Congress often deploys “standardized formula[s]” or checklists, like the one found in paragraph (f )(1), that “are by their nature over- and under-inclusive.” Ransom v. FIA Card Services, N. A., 562 U. S. 61, 78 (2011). And because Congress may rationally prefer these approaches for various reasons, including their ease of administration, this Court has long held that we will not second-guess them merely because they may produce some “oddit[ies]” or “anomalies.” Ibid.; see Rodriguez v. United States, 480 U. S. 522, 526 (1987) (per curiam) (“Deciding what competing values will or will not be sacrificed to the achievement of a particular objective is the very essence of legislative choice”).⁸

 If, as the government supposes, a seemingly anomalous result alone could unsettle a statute, it would face its own troubles, too. Under its implicit distribution theory, an individual who previously committed a nonviolent offense and received a sentence longer than 13 months (i.e., a 3-point offense) is categorically ineligible for relief. Meanwhile, an individual who committed the same crime but received a sentence equal to or one day less than 13 months (i.e., a 2-point offense) thanks to a slightly more lenient sentencing judge remains eligible for relief. Rather than “unerringly” enable the safety valve to “separat[e] more serious prior offenders from less serious ones,” ante, at 21, the government’s approach thus leaves much to happenstance and luck—an anomalous result indeed.

 Return, then, to our actual absurdity doctrine and consider the government’s argument in its light. The government worries that respecting paragraph (f )(1) as written would treat “more serious” offenders too leniently. But in doing so, the government ignores what follows. A defendant who satisfies paragraph (f )(1) must still go on to satisfy paragraphs (f )(2)–(5). And those provisions collectively operate to deny relief to virtually anyone whose current offense involves any trace of violence.

 Even if a “more serious” offender could somehow thread  his way through all those needles, too, another would await. The safety valve instructs a sentencing court to fashion a sentence “pursuant to [the] guidelines.” §3553(f ). The guidelines expressly account for a defendant’s criminal history, and few would accuse them of leniency toward those with a history of serious offenses. In fact, defendants with significant criminal histories often wind up with a recommended guidelines sentence higher than the otherwise-applicable mandatory minimum. See Brief for National Association of Federal Defenders as Amicus Curiae 7–8. Sentencing courts may have the discretion to vary or depart from the guidelines’ recommended ranges. But Congress could have rationally trusted courts to exercise that discretion with an appreciation for the fact that individuals with serious criminal histories—such as the government’s hypothetical defendant with many prior three-point violent offenses, see ante, at 21–22; Brief for United States 23—warrant equally serious sentences. So, looking to the law as a whole (as we must) and appreciating that Congress often legislates using standardized formulas or checklists that may be over- and under-inclusive (again, as we must), there is nothing approaching an absurdity that might license us to rewrite the First Step Act.⁹

 In a final effort to bolster the government’s case, the Court professes an entirely different concern of its own. It claims to worry that the natural reading of the law would  sometimes be too harsh in operation. Holding a defendant ineligible for safety-valve relief based on offenses that score “zero points,” we are told, would be “out-of-sync” with the law’s purpose. Ante, at 20. But there is nothing absurd here either. Subparagraph (A) provides that defendants are eligible for relief as long as their past convictions do not yield more than four criminal history points—a calculation that, as we have seen, does not include points associated with old crimes and certain other offenses. Subparagraphs (B) and (C) provide that other defendants with more than four criminal history points are eligible for relief too as long as they don’t have anywhere in their past a serious (3-point) offense and a weighty (2-point) violent offense—even if those offenses are (say) too old to contribute to their criminal history scores. So whatever unfairness the Court may perceive in one part of the safety valve (here, subparagraphs (B) and (C)) is diminished when considered in light of another (here, subparagraph (A)). Some might prefer a different arrangement, but the one Congress ordained is hardly absurd.

 If any law demonstrates the wisdom of our usual rules against elevating policy appeals over plain text, it is this one. Under the ordinary meaning of the statute, it is possible some “more serious” offenders may make it past paragraph (f )(1), and perhaps even end up receiving an individualized sentence under guidelines that hardly exhibit solicitude for those with “more serious” criminal histories. Under the implicit distribution theory, in contrast, the availability of individualized sentencing may depend on the happenstance of one extra day in prison. In the end, attempting to pick between these two outcomes proves nothing more than the futility of the exercise. However artfully the government frames its dissatisfaction with the text of the statute, we have neither the institutional competence nor the constitutional mandate “to assess the relative merits of different approaches” Congress could have taken. Hartford Underwriters Ins. Co. v. Union Planters Bank, N. A., 530 U. S. 1, 13 (2000). Our role is a more modest one: “[W]hen the statute’s language is plain” and “the disposition required by the text is not absurd,” “the sole function of the courts . . . is to enforce it according to its terms.” Id., at 6 (internal quotation marks omitted). Because that is undoubtedly the case here, we must apply the safety valve as written.

VI

 As I see it, the government hasn’t come close to supplying a lawful basis for departing from the law’s ordinary meaning. Suppose, though, at the end of this long march through its inventive theories you remain unsure. Suppose you are left with a reasonable doubt about whether Mr. Pulsifer or the government has the better reading of the law. In circumstances like that, another rule of construction supplies an answer. It is lenity.

 The rule of lenity “is perhaps not much less old than construction itself.” United States v. Wiltberger, 5 Wheat. 76, 95 (1820) (Marshall, C. J.); see Wooden v. United States, 595 U. S. 360, 388 (2022) (Gorsuch, J., concurring in judgment) (citing The Adventure, 1 F. Cas. 202, 204 (No. 93) (CC Va. 1812) (Marshall, C. J.)). It requires courts to interpret ambiguous “penal laws,” including those concerning sentencing, in favor of liberty, not punishment. Wiltberger, 5 Wheat., at 95; United States v. Batchelder, 442 U. S. 114, 121 (1979); Bifulco v. United States, 447 U. S. 381, 387 (1980).

 This rule enforces weighty constitutional values. Courts construe ambiguous penal laws with lenity because a free nation operates against a background presumption of individual liberty. See Wooden, 595 U. S., at 391–392 (opinion of Gorsuch, J.). We resolve doubts about a criminal law’s reach in favor of lenity, too, because in our federal government only the people’s elected representatives, not their  judges, are vested with the power to “define a crime, and ordain its punishment.” Wiltberger, 5 Wheat., at 95; accord, Crandon v. United States, 494 U. S. 152, 158 (1990); Wooden, 595 U. S., at 391–392 (opinion of Gorsuch, J.).

 Lenity protects vital due process interests, as well, by ensuring individuals fair notice of the consequences of their actions. United States v. Lanier, 520 U. S. 259, 266 (1997); see McBoyle v. United States, 283 U. S. 25, 27 (1931); Wooden, 595 U. S., at 389–391 (opinion of Gorsuch, J.). And lenity performs still further work, guarding against the possibility that judges might condemn unpopular individuals to punishment on the strength of their own views about common sense, good public policy, or “no more than a guess as to what Congress intended.” Ladner v. United States, 358 U. S. 169, 178 (1958).

 So suppose you thought a reasonable doubt remained about how best to construe the First Step Act. In those circumstances, the answer cannot be to adopt an interpretation that restricts safety-valve relief to thousands more individuals. The only permissible answer is one that favors liberty.

VII

 Today, the Court does not hedge its doubts in favor of liberty. Instead, it endorses the government’s implicit distribution theory and elevates it over the law’s ordinary and most natural meaning.

 It is a regrettable choice that requires us to abandon one principle of statutory interpretation after another. We must read words into the law; we must delete others. We must ignore Congress’s use of a construction that tends to avoid, not invite, questions about implicit distribution. We must dismiss Congress’s variations in usage as sloppy mistakes. Never mind that Congress distributed phrases expressly when it wanted them to repeat in the safety valve.  Never mind that Congress used “or” when it sought an efficient way to hinge eligibility for relief based on a single characteristic. We must then read even more words yet into the law to manufacture a superfluity problem that does not exist. We must elevate unexpressed congressional purposes over statutory text. Finally, rather than resolve any reasonable doubt about statutory meaning in favor of the individual, we must prefer a more punitive theory the government only recently engineered.

 Today, the Court indulges each of these moves. All to what end? To deny some individuals a chance—just a chance—at relief from mandatory minimums and a sentence that fits them and their circumstances. It is a chance Congress promised in the First Step Act, and it is a promise this Court should have honored. Respectfully, I dissent.