Harvard Law Review Harvard Law Review Harvard Law Review

Harvard Law Review Forum

Avoiding Constitutional Questions Versus Avoiding Unconstitutionality

In this month’s issue, Professor Neal Katyal and Thomas Schmidt join a distinguished group of lawyers and judges who criticize the “canon of constitutional avoidance” — the idea that courts should try to interpret statutes so as to avoid raising difficult questions of constitutional law.1×1. Neal Kumar Katyal & Thomas P. Schmidt, Active Avoidance: The Modern Supreme Court and Legal Change, 128 Harv. L. Rev. 2109 (2015); cf. Neal Kumar Katyal, Judges as Advicegivers, 50 Stan. L. Rev. 1709, 1815 (1998) (alluding to problems with the canon). Although the Supreme Court has described this canon as a “settled policy,”2×2. Gomez v. United States, 490 U.S. 858, 864 (1989). the canon exists more by dint of repetition than by force of argument. Its critics include the most eminent circuit judge of the last generation,3×3. See Henry J. Friendly, Mr. Justice Frankfurter and the Reading of Statutes, in Benchmarks 196, 209–12 (1967). two of the most eminent circuit judges of the present generation,4×4. See Frank H. Easterbrook, Do Liberals and Conservatives Differ in Judicial Activism?, 73 U. Colo. L. Rev. 1401, 1405–06 (2002); Richard A. Posner, Statutory Interpretation — In the Classroom and in the Courtroom, 50 U. Chi. L. Rev. 800, 815–16 (1983). and a host of thoughtful scholars.5×5. Early entrants include Lawrence C. Marshall, Divesting the Courts: Breaking the Judicial Monopoly on Constitutional Interpretation, 66 Chi.-Kent L. Rev. 481, 483–92 (1990); Frederick Schauer, Ashwander Revisited, 1995 Sup. Ct. Rev. 71; Lisa A. Kloppenberg, Avoiding Serious Constitutional Doubts: The Supreme Court’s Construction of Statutes Raising Free Speech Concerns, 30 U.C. Davis L. Rev. 1 (1996); John Copeland Nagle, Delaware & Hudson Revisited, 72 Notre Dame L. Rev. 1495 (1997); and William K. Kelley, Avoiding Constitutional Questions as a Three-Branch Problem, 86 Cornell L. Rev. 831 (2001).

All three of the judges just mentioned, and many of the scholars, have criticized only the canon that favors avoiding serious constitutional questions. They have not objected to a separate canon that favors avoiding actual unconstitutionality — the longstanding principle that courts should not lightly interpret a statute in a way that makes it unconstitutional if some other interpretation is available.6×6. See Friendly, supra note 3, at 210 (calling the latter canon a “principle of unquestionable validity”); Easterbrook, supra note 4, at 1405–06; Posner, supra note 4, at 814–15.

Katyal and Schmidt downplay this distinction. In part, that reflects their assessment of current practice: they agree with Professor Adrian Vermeule that the canon about avoiding actual unconstitutionality (which Vermeule dubbed “classical avoidance” to reflect its historical pedigree) “has been mostly superseded” by the canon about avoiding even constitutional questions (which Vermeule dubbed “modern avoidance”).7×7. Katyal & Schmidt, supra note 1, at 2117; see also Adrian Vermeule, Saving Constructions, 85 Geo. L.J. 1945, 1949 (1997) (introducing the labels “classical avoidance” and “modern avoidance,” and adding that the latter has “supplanted” the former). But Katyal and Schmidt see little reason to resurrect the distinction. In their telling, both canons are variations on the same theme, and both are subject to the same objections: the canons lead to bad constitutional doctrine (because they enable judges to articulate new constitutional principles without the discipline that allegedly comes from “actually having to strike down a law”8×8. Katyal & Schmidt, supra note 1, at 2123. ), and they also lead to bad interpretations of statutes (because they encourage judges to distort statutes in order to avoid whatever constitutional difficulties the judges have identified).

I come to this topic having been persuaded by the prior critics,9×9. See Caleb Nelson, Statutory Interpretation 146–49 (2011). so this Response defends the distinction that Katyal and Schmidt downplay. To begin with, I do not think that the canon about avoiding unconstitutionality is as rare as Vermeule suggested. Admittedly, courts often do conflate it with the canon about avoiding constitutional questions,10×10. See, e.g., Edward J. DeBartolo Corp. v. Fla. Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 575 (1988); see also Friendly, supra note 3, at 210 (observing that these two principles of statutory interpretation are “sometimes not sufficiently distinguished”). But cf. Rust v. Sullivan, 500 U.S. 173, 190–91 (1991) (describing the two canons as related but distinct principles). and the Supreme Court sometimes uses a single formulation for both.11×11. See, e.g., Office of Senator Mark Dayton v. Hanson, 550 U.S. 511, 514 (2007) (referring to “our established practice of interpreting statutes to avoid constitutional difficulties”). But the lower federal courts and (especially) the state courts continue to refer specifically to the canon about avoiding unconstitutionality.12×12. See, e.g., Galarza v. Szalczyk, 745 F.3d 634, 643 (3d Cir. 2014); License Acquisitions, LLC v. Debary Real Estate Holdings, LLC, 155 So. 3d 1137, 1146–47 (Fla. 2014). Likewise, the modern Supreme Court has indicated that the canon about avoiding unconstitutionality takes precedence over the canon about avoiding constitutional questions13×13. Edmond v. United States, 520 U.S. 651, 658 (1997) (concluding that judges must not interpret a federal statute “in a manner that would render it clearly unconstitutional” if “another reasonable interpretation [is] available,” even if that other interpretation entails reading a different federal statute to raise “a constitutional question”). — a conclusion that reflects not only an ongoing distinction between the two canons but also an implicit assessment of their respective justifications.

In any event, to the extent that current doctrine does not adequately differentiate these canons, Katyal and Schmidt provide reasons to do so. As we shall see, their arguments about statutory interpretation (discussed in Part I of this Response) and their concerns about sloppy constitutional reasoning (discussed in Part II) both apply more readily to the canon about avoiding constitutional questions than the canon about avoiding actual unconstitutionality. Ultimately, Part III of this Response advocates abandoning the “questions” canon even more thoroughly than Katyal and Schmidt propose. But a version of the canon about avoiding actual unconstitutionality strikes me as a more plausible tool of statutory interpretation, and I am not as concerned as Katyal and Schmidt about its effects on constitutional doctrine.


Efforts to evaluate canons of statutory interpretation are hampered by the fact that different people have different views about the goals of interpretation. Understandably, Katyal and Schmidt try to cast their arguments in terms that will resonate with a wide variety of readers. Thus, they speak of the “best” reading of a statute without specifying what makes one reading better than another, and they define the “distortion” introduced by avoidance as “[the] space . . . between the best reading according to the interpreter’s ideal method and the avoidance-compelled reading.”14×14. Katyal & Schmidt, supra note 1, at 2116. They acknowledge, however, that one cannot really assess a canon while remaining completely agnostic about the functions that canons are supposed to serve.15×15. See id. at 2116 n.23. Katyal and Schmidt’s seemingly neutral definition of “distortion” does not escape this problem. After all, it assumes that “the interpreter’s ideal method” does not include the avoidance canon — an assumption that supporters of the canon might resist. Cf. Ernest A. Young, Constitutional Avoidance, Resistance Norms, and the Preservation of Judicial Review, 78 Tex. L. Rev. 1549, 1591 (2000) (complaining that “the critics’ conception of an ‘otherwise preferred’ . . . reading is artificial,” and arguing that constitutional concerns can legitimately affect statutory interpretation). As their argument proceeds, Katyal and Schmidt therefore seem to associate “distortions” of a statute with departures from the meaning intended by Congress.16×16. See, e.g., Katyal & Schmidt, supra note 1, at 2127 (“[T]he more Congress’s words have been distorted, the less likely Congress can be presumed to intend the distortion . . . .”); id. at 2128 (describing “distortion” as “thwart[ing] congressional intent”).

That may sound controversial. Textualists often speak as if collective entities like legislatures do not really have intentions.17×17. See, e.g., Antonin Scalia & Bryan A. Garner, Reading Law 392–93 (2012). And even if one gets past that objection, no one thinks that the sole goal of statutory interpretation is to capture the subjective intentions of members of the enacting legislature. Because of the importance of other goals (like enabling people in the real world to have fair notice of what statutory language will be understood to mean), everyone acknowledges that a particular statute might mean something other than what members of the enacting legislature subjectively intended it to mean.

But while such breakdowns in communication are inevitable, courts should not go out of their way to foster them. Even textualists, who want courts to apply their normal principles of interpretation without making ad hoc assessments of legislative intent, favor principles of interpretation that will promote successful communication from the legislature to the courts.18×18. For general thoughts on the relationship between textualism and legislative intent, see Caleb Nelson, What Is Textualism?, 91 Va. L. Rev. 347 (2005). In my view, a canon that impedes such communication — that is, a canon that systematically reduces the correlation between the meanings that courts ascribe to statutes and the meanings that members of the enacting legislature collectively intended to convey — is not a canon that courts should use.

I agree with Katyal and Schmidt that the canon about avoiding constitutional questions probably fails this test. But a version of the canon about avoiding unconstitutionality may pass it.


There are at least two possible formulations of the canon about avoiding constitutional questions. On the broader formulation, the canon amounts to a general principle of abstention: if one reading of a statute would force courts to confront hard constitutional questions, courts should prefer an alternative reading that lets them duck those questions. On the narrower formulation, the canon simply tells courts to avoid interpretations that would raise hard questions about the constitutionality of the statute itself.

To see the difference, think of statutes that go precisely as far as the Constitution permits. For instance, Rhode Island’s long-arm statute makes defendants “amenable to suit in Rhode Island in every case not contrary to the provisions of the constitution or laws of the United States.”19×19. R.I. Gen. Laws § 9-5-33(a) (2014). Courts applying this statute will regularly have to confront hard questions of constitutional law, but the constitutionality of the statute itself will not be in doubt. The same is true of many other statutes that incorporate constitutional concepts by reference.20×20. For one of the myriad examples in federal law, see 18 U.S.C. § 3731 (2012) (“In a criminal case an appeal by the United States shall lie to a court of appeals from a decision, judgment, or order of a district court dismissing an indictment or information or granting a new trial after verdict or judgment, . . . except that no appeal shall lie where the double jeopardy clause of the United States Constitution prohibits further prosecution.”); see also Vermeule, supra note 7, at 1965 n.100 (citing cases that read particular federal statutes “to reach to the limits of Congress’s constitutional power over interstate commerce”).

The familiarity of such statutes is a strike against the broader formulation of the canon about avoiding constitutional questions. Even if legislatures generally try to enact statutes that stay within constitutional bounds, legislatures apparently do not have a general practice of trying to save courts from addressing what those bounds are. And if the broader formulation of the canon does not reflect accurate generalizations about legislative behavior, it is not a good guide to the intended meaning of statutes.

But the fact that legislatures sometimes use their powers to the fullest does not necessarily defeat the narrower formulation of the canon. From the standpoint of legislative intent, here is the key question: If a statute lends itself to two possible interpretations, but one of those interpretations would raise serious questions about the statute’s own constitutionality, can interpreters infer that members of the enacting legislature probably had the other interpretation in mind?

My answer is: Maybe sometimes, but an interpreter should do a lot more work before drawing this inference. First, the interpreter should ask whether members of the enacting legislature would have been aware of the reasons for doubting the constitutionality of the one interpretation. If so, the interpreter should ask whether members of the enacting legislature would have cared. Are the doubts that the interpreter has identified the sorts of doubts that would have troubled legislators, or would legislators have been likely to press ahead regardless (hoping, perhaps, that courts would resolve the doubts in favor of the statute’s constitutionality)?

Professor Frederick Schauer has suggested that this analysis will rarely support avoidance. In his words, “there is no evidence whatsoever that members of Congress are risk-averse about the possibility that legislation they believe to be wise policy will be invalidated by the courts.”21×21. Schauer, supra note 5, at 92. Given their incentives, moreover, “[o]ne would expect them to err on the side of assuming constitutionality under conditions of uncertainty about what the courts are likely to do.”22×22. Id. at 92–93.

That surmise strikes me as a bit too strong. Some members of Congress might have an ethic of avoiding certain kinds of constitutional doubts. And even if legislators care primarily about wise policy, there are sometimes policy reasons for legislators to favor a proposal that will not be subject to constitutional attack over a proposal that will be (especially if the difference between the two proposals is not crucial to what the legislators want to accomplish). In some circumstances, then, the fact that a particular interpretation raises constitutional doubts might indeed suggest that members of the enacting legislature had an alternative interpretation in mind.

Still, identifying those circumstances requires fine-grained information of the sort that is hard to “canonize” — that is, to reduce to a general rule of thumb that remains reasonably accurate. As a result, I am inclined to think that arguments about the implications of constitutional doubts for statutory interpretation are best advanced doubt by doubt and statute by statute. That puts me in Katyal and Schmidt’s camp: the generalization that legislatures try to avoid enacting statutes whose constitutionality might be doubted is unlikely to be true often enough to justify having a canon to that effect.23×23. It might be possible to identify some discrete areas of constitutional law in which legislators routinely try not to push the envelope. For instance, Congress might try not to be very aggressive about legislating retroactively, or about foreclosing judicial review of agency action. With respect to those issues, though, we can and do have targeted canons that reflect the legislature’s usual practices.

Of course, even if a general canon about avoiding constitutional questions is not a reliable guide to the intended meaning of statutory language, courts could still use such a canon as a tiebreaker when their ordinary tools for identifying a statute’s intended meaning leave them in equipoise between two readings. That is how the rule of lenity works,24×24. Katyal & Schmidt, supra note 1, at 2162–63; see also Barber v. Thomas, 130 S. Ct. 2499, 2508–09 (2010) (“[T]he rule of lenity only applies if, after considering text, structure, history, and purpose, there remains a ‘grievous ambiguity or uncertainty in the statute’ such that the Court must simply ‘guess as to what Congress intended.’” (citations omitted) (quoting Muscarello v. United States, 524 U.S. 125, 139 (1998) (internal quotation marks omitted); Bifulco v. United States, 447 U.S. 381, 387 (1980))). and a canon about avoiding constitutional questions might serve a similar function.25×25. Cf. Nelson, supra note 18, at 393–98 (discussing the role of “normative” canons like the rule of lenity and the canon about avoiding constitutional questions). In my view, though, canons that are not accurate generalizations about legislative intent do not deserve the same weight in the interpretive process as canons that are.


While I am skeptical that legislatures try to steer clear of all constitutional lines, I am prepared to accept the generalization that legislatures try to avoid enacting statutes that will actually be held unconstitutional. If two readings of a statute are both fairly possible, and if members of the enacting legislature would have thought that they lacked the authority to establish Interpretation #1 (or that courts probably would so hold), that fact might normally be some evidence that members of the enacting legislature had Interpretation #2 in mind instead.

At the very least, this inference about legislative intent seems more plausible than the idea that legislators routinely try to avoid mere constitutional questions. In support of this distinction, Professor Einer Elhauge points to state legislatures’ choices about which canons to codify. A number of state legislatures have explicitly endorsed the presumption that when they enact a statute, “[c]ompliance with the Constitutions of the state and of the United States is intended.”26×26. Iowa Code § 4.4 (2013). For nearly identical provisions, see Colo. Rev. Stat. § 2-4-201 (2014); Minn. Stat. § 645.17 (2014); N.D. Cent. Code § 1-02-38 (2008); Ohio Rev. Code Ann. § 1.47 (LexisNexis 2009); 1 Pa. Cons. Stat. § 1922 (2008); Tex. Gov’t Code Ann. § 311.021 (West 2013); see also N.M. Stat. Ann. § 12-2A-18 (2014). These citations are collected in Einer Elhauge, Statutory Default Rules 354 n.37 (2008). By contrast, not a single state legislature has told courts to presume that it intends to avoid even constitutional doubts.27×27. See Elhauge, supra note 26, at 237.

Admittedly, as Professor Schauer noted two decades ago, the idea that legislatures shy away from enacting unconstitutional statutes rests on intuitions that have not been tested empirically.28×28. See Schauer, supra note 5, at 92 & n.61. Professors Abbe Gluck and Lisa Bressman recently took a step in the right direction by sending a survey to congressional staffers, focusing on people who work as counsel to congressional committees or in the Offices of the House and Senate Legislative Counsel. As I understand the results, though, the portion of the survey that dealt with constitutional issues did not yield much relevant information. See Abbe R. Gluck & Lisa Schultz Bressman, Statutory Interpretation from the Inside — An Empirical Study of Congressional Drafting, Delegation, and the Canons: Part I, 65 Stan. L. Rev. 901, 947–48 (2013); cf. Donald G. Morgan, Congress and the Constitution app. A (1966) (reporting an earlier but less detailed survey of members of Congress themselves); Bruce G. Peabody, Congressional Constitutional Interpretation and the Courts: A Preliminary Inquiry into Legislative Attitudes, 1959–2001, 29 Law & Soc. Inquiry 127, 146–56 (2004) (updating Morgan’s survey). So far as I know, scholars have not systematically examined how either the Constitution or the courts’ constitutional doctrines might influence drafting choices within Congress. I note, however, that a scholar with extensive experience on Capitol Hill accepts the canon about avoiding unconstitutionality but criticizes the canon about avoiding constitutional questions. See Edward O. Correia, A Legislative Conception of Legislative Supremacy, 42 Case W. Res. L. Rev. 1129, 1177–80 (1992). Schauer’s own intuitions led him to suggest that “Congress is rarely concerned with the fact of unconstitutionality independent of the likelihood that an Act of Congress will be overturned by the courts.”29×29. Schauer, supra note 5, at 92. That premise, though, actually makes the canon about avoiding unconstitutionality more tractable than it would otherwise be: instead of needing information about how members of Congress themselves understand the Constitution, we can apply the canon primarily with reference to prevailing judicial doctrines.

Of course, predicting how a particular statutory provision might fare in court is a probabilistic enterprise. Thus, Katyal and Schmidt may be correct that the canon about avoiding unconstitutionality lies on the same spectrum as a canon about avoiding constitutional doubts. But as doubts grow into probabilities, the canon becomes an increasingly plausible guide to legislative intent.

The fact that prevailing judicial doctrines can change over time raises a complication. If members of the enacting Congress would have thought that both possible interpretations of a statute were perfectly constitutional, the fact that courts now consider one of those interpretations unconstitutional does not really suggest that members of Congress probably intended the other interpretation.30×30. See Katyal & Schmidt, supra note 1, at 2126–27. My own view, then, is that courts should apply the canon about avoiding unconstitutionality mostly with reference to the constitutional doctrines that were familiar when a statute was enacted.31×31. See Robert W. Scheef, Temporal Dynamics in Statutory Interpretation: Courts, Congress, and the Canon of Constitutional Avoidance, 64 U. Pitt. L. Rev. 529, 531 (2003) (calling attention to this issue and urging courts to “construe congressional intent within its temporal context — in light of the constitutional law as it stood at the time Congress enacted the statute”). I say “mostly” because more recently articulated constitutional doctrines can serve as tiebreakers: if a court is in equipoise about whether the legislature intended Interpretation #1 or Interpretation #2, the fact that the court would now hold Interpretation #1 unconstitutional can be a reason for the court to pick Interpretation #2 instead.

Supporters of the canon sometimes respond that even when the canon is not a plausible guide to the intended meaning of a specific statute, the canon accurately reflects a different sort of intent — Congress’s intentions about the principles of interpretation that courts should use. Suppose that a provision lends itself to two possible interpretations that both would have been considered constitutional at the time of enactment, but modern courts now regard Interpretation #1 as unconstitutional. Some people suggest that in this situation, even if members of the enacting legislature had actually intended to establish Interpretation #1, they themselves would want courts to adopt Interpretation #2 rather than holding the statute unconstitutional.32×32. Even if this argument works (and I do not think it does), it is specific to the canon about avoiding unconstitutionality. For reasons explained by Judge Friendly, the argument does not support the canon about avoiding mere constitutional doubts. See Friendly, supra note 3, at 210.

This particular defense of the canon goes too far for me. Depending on what Interpretation #1 and Interpretation #2 are, a legislature that actually intended to establish Interpretation #1 might consider Interpretation #2 worse than no statute at all. Often, moreover, that is not really the correct comparison. When Interpretation #1 makes the statute only partially unconstitutional, the presumption of severability would encourage courts to apply the statute where it is constitutional and to disregard the statute only in its unconstitutional aspects or applications. A legislature that had intended to establish Interpretation #1 might well prefer that outcome to the results produced by Interpretation #2.33×33. Professor Vermeule apparently would reject this point, but only because he mistakenly assumes that what I am calling “Interpretation #2” always amounts to the same thing as “Interpretation #1 minus the applications that would be unconstitutional.” See Vermeule, supra note 7, at 1957 n.69 (asserting that the canon about avoiding unconstitutionality “is in effect identical to severance conducted after a constitutional ruling”). That is sometimes true, but not always. Statutory language often lends itself to two discrete interpretations, neither of which is wholly subsumed by the other. And even when one interpretation is just a subset of the other, the narrower interpretation may be significantly narrower than the Constitution requires; in other words, the narrower interpretation might not encompass all of the applications of the broader provision that the Constitution would permit. In some contexts, moreover, a particular interpretation of a statute might be unconstitutional because it is too narrow rather than too broad. Because of the variety of both statutory language and constitutional problems, the canon about avoiding unconstitutionality is not functionally identical to the presumption of severability.

In sum, I think that the canon about avoiding unconstitutionality should focus on the constitutional doctrines that were familiar when the relevant statute was enacted. Subject to that nuance, though, the canon strikes me as a plausible guide to the statute’s intended meaning. While the generalization that the canon reflects is not always true, and while I do not think that the canon is strong enough to justify strained interpretations of statutory language,34×34. Cf. Clark v. Martinez, 543 U.S. 371, 400 (2005) (Thomas, J., dissenting) (criticizing “the Court’s aggressive application of modern constitutional avoidance doctrine”); Katyal & Schmidt, supra note 1, at 2163 (discussing the concept of interpretations that are “fairly possible” (quoting United States v. Jin Fuey Moy, 241 U.S. 394, 401 (1916))). the canon deserves some weight in the interpretive process.


In addition to arguing that canons of constitutional avoidance lead to bad interpretations of statutes (because the canons drive courts away from what Congress intended), Katyal and Schmidt also argue that they lead to bad constitutional doctrine.35×35. See Katyal & Schmidt, supra note 1, at 2122–26. Again, though, this argument is better directed against the canon about avoiding constitutional questions than the canon about avoiding actual unconstitutionality.

As Katyal and Schmidt observe, the canon about avoiding constitutional questions may lead courts to engage in “sloppy and cursory constitutional reasoning.”36×36. Id. at 2112. Other scholars agree. See, e.g., Schauer, supra note 5, at 89–90. To be sure, before invoking the canon, courts are supposed to analyze the relevant constitutional issues enough to know that one interpretation of the statute raises grave constitutional concerns. In the Supreme Court’s words, “those who invoke the doctrine must believe that the alternative is a serious likelihood that the statute will be held unconstitutional.”37×37. Almendarez-Torres v. United States, 523 U.S. 224, 238 (1998). But courts are not actually supposed to answer the specific constitutional questions that they use the canon to avoid. In practice, opinions invoking the canon may therefore rely on handwaving rather than rigorous analysis.

According to Katyal and Schmidt, moreover, the handwaving sometimes ends up having precedential effect in later cases.38×38. See Katyal & Schmidt, supra note 1, at 2135–36 (discussing the role that the Court’s opinion in Northwest Austin Municipal Utility District No. One v. Holder, 557 U.S. 193 (2009), played in Shelby County v. Holder, 133 S. Ct. 2612 (2013)). The scope of that problem is not clear. Indeed, Adrian Vermeule has previously made an opposing assertion: “The case law is rife with constitutional questions that the Court has avoided by construction, only later to hold, when forced to confront the question under a different statute, that the constitutional claim should not prevail.”39×39. Vermeule, supra note 7, at 1960. But Vermeule’s examples are not really apt.40×40. For instance, Vermeule’s lead example involved the alleged contrast between NLRB v. Catholic Bishop of Chicago, 440 U.S. 490 (1979), and Tony & Susan Alamo Foundation v. Secretary of Labor, 471 U.S. 290 (1985). But the constitutional question that the Court ducked in Catholic Bishop (whether the Religion Clauses of the First Amendment prevent the National Labor Relations Act from reaching teachers at church-operated schools) is not the same question that the Court answered in Alamo Foundation (which held that the Religion Clauses did not prevent the federal minimum-wage statute from reaching commercial enterprises that a religious foundation ran to provide jobs for people needing rehabilitation). In Catholic Bishop, the Court’s constitutional concerns centered on “the . . . unique role of the teacher in fulfilling the mission of a church-operated school,” 440 U.S. at 501, and the danger that government supervision of collective bargaining between teachers and church-operated schools would lead to “entanglement with the religious mission of the school,” id. at 502. Those concerns were not present in Alamo Foundation. See Alamo Foundation, 471 U.S. at 305 & n.31 (distinguishing Catholic Bishop).
Of course, I do not defend the majority’s performance in Catholic Bishop. In the name of the canon about avoiding constitutional questions, the majority simply invented an exception to the National Labor Relations Act for teachers in church-operated schools. See Catholic Bishop, 440 U.S. at 499–507.
While we lack systematic research on the horizontal, vertical, and interbranch precedential effects of avoidance decisions, it is possible that the canon about avoiding constitutional questions is bad for constitutional doctrine.

I am not persuaded, however, that the canon about avoiding actual unconstitutionality suffers from the same problem. Katyal and Schmidt are particularly concerned about what they call “generative avoidance” — the possibility that courts will apply the avoidance canon with reference to constitutional doctrines that the courts have not previously articulated.41×41. Katyal & Schmidt, supra note 1, at 2112, 2118, 2122–26. My preferred version of the canon about avoiding unconstitutionality is not subject to those concerns: if courts applied the canon with reference to the constitutional doctrines that prevailed when the relevant statute was enacted,42×42. See supra text accompanying notes 30–31, pp. 337–38. the canon would not invite “generative avoidance.” But Katyal and Schmidt’s concerns may be unfounded anyway.

Katyal and Schmidt worry that when the Supreme Court articulates new constitutional doctrine in the context of interpreting a statute, and when the Court is using that doctrine simply to favor one interpretation of the statute over the other, the Court’s statements about the doctrine are “nearly costless for [the] [C]ourt” because “no law is struck down as a result.”43×43. Katyal & Schmidt, supra note 1, at 2126. By contrast, when the Supreme Court is articulating new principles of constitutional law in cases where they will have their “full logical consequence” (that is, cases where a statute is being “invalidat[ed]”),44×44. Id. at 2123. For appropriate cautions about saying that judicial decisions “invalidate” statutes, see John Harrison, Severability, Remedies, and Constitutional Adjudication, 83 Geo. Wash. L. Rev. 56, 81–90 (2014). the Court will expect more “public scrutiny and criticism” of whatever it says, and it will be correspondingly more careful.45×45. Katyal & Schmidt, supra note 1, at 2126. The fact that the Court is seeing at least one concrete application of its new constitutional doctrine may also give the Court a better sense of the doctrine’s practical consequences.46×46. See id. at 2123–26. Katyal and Schmidt conclude that the content of our constitutional doctrine will be better if the Court develops that doctrine in cases about the validity of a statute rather than just the meaning of a statute.

That argument, however, rests on a series of questionable empirical assumptions. Start with the effects of outside scrutiny. Perhaps the Court does face more scrutiny when the Court holds a statute unconstitutional than when the Court simply says that a statute would be unconstitutional if it were interpreted to establish a directive that the Court does not actually interpret it to establish. And perhaps “public and academic scrutiny . . . is one of [our] most efficacious protections against judicial overreach.”47×47. Id. at 2134. But that is simply to say that we do not have much protection against judicial overreach. With the possible exception of a few highly politicized cases each Term, I doubt that either the Supreme Court as an institution or the Justices as individuals feel significant costs when the Court holds a statute unconstitutional. Our primary guarantee that the Justices will do their jobs as well as they can is not cost internalization but the Justices’ own good faith.

In any event, lower courts decide thousands more cases each year than the Supreme Court, and their decisions do not receive anything like the scrutiny that Katyal and Schmidt have in mind. If the canon about avoiding unconstitutionality would otherwise be a useful tool of statutory interpretation, I very much doubt that considerations unique to the Supreme Court should make us want to abandon it.

Aside from outside scrutiny, perhaps one could argue that judges feel a greater internal sense of responsibility when they are reviewing a statute’s constitutionality than when they are analyzing constitutional issues as a prelude to interpreting the statute. But to the extent that judges still think of judicial review in the terms suggested by Marbury v. Madison,48×48. 5 U.S. (1 Cranch) 137 (1803). these two occasions for constitutional analysis are not really so different. According to Marbury, whether judges are engaging in judicial review or statutory interpretation, they are trying to identify the rules of decision for particular cases.49×49. See Harrison, supra note 44, at 86 (explaining that Marbury rested on “the judicial duty to decide cases according to law”). To appreciate Marbury’s account of judicial review, suppose that someone brings a case in a court of competent jurisdiction, and one of the parties invokes a federal statute that purports to supply a rule of decision. Ordinarily, the court is supposed to decide the case accordingly. But statutes are not the only form of law that courts recognize; the Constitution is also law of this sort, and it bears on the rules of decision that legislatures have power to establish. As a result, when judges are trying to identify a rule of decision for a particular case, they are not supposed to “close their eyes on the constitution, and see only the [statute].” Marbury, 5 U.S. (1 Cranch) at 178. If they conclude that the legislature lacks the power to establish the rule of decision that the statute purports to supply, then they will not apply that rule in the case that they are adjudicating. That is “judicial review.” From a judge’s standpoint, moreover, there may not be a vast psychological difference between (1) saying that a purported rule of decision is not applicable to this case because it is unconstitutional and (2) saying that a purported rule of decision is not applicable to this case because its unconstitutionality leads the judge to think that Congress intended something else.50×50. Admittedly, Justice Holmes may have seen more of a difference than I perceive. In his separate opinion in Blodgett v. Holden, 275 U.S. 142 (1927), he explained the canon about avoiding unconstitutionality partly on the ground that “to declare an Act of Congress unconstitutional . . . is the gravest and most delicate duty that this Court is called on to perform,” id. at 147–48 (opinion of Holmes, J.), quoted in Katyal & Schmidt, supra note 1, at 2123, 2124 n.53. At the time that Justice Holmes was writing, though, prominent politicians and others were questioning the very institution of judicial review. See generally William G. Ross, A Muted Fury: Populists, Progressives, and Labor Unions Confront the Courts, 1890–1937 (1994) (canvassing Progressive-era attacks and proposals for reform). Professor Schauer is surely correct that judicial review now seems “less exceptional.” Schauer, supra note 5, at 96; see also, e.g., Webster v. Doe, 486 U.S. 592, 618 (1988) (Scalia, J., dissenting) (pointing out that statutory rights can be more important to claimants than constitutional rights).

Nor is it obvious that judges will consistently have less information about the practical consequences of a proposed constitutional doctrine when the judges are applying the canon about avoiding unconstitutionality than when they are being asked to hold a statute unconstitutional. While there may be some systematic difference along these lines, my bet is that it is relatively small. Indeed, I suspect that both types of arguments often come up in the same case: a party might ask the court to hold either that a statute is unconstitutional (under some doctrine that the party wants the court to recognize) or that the statute should be interpreted so as to preserve its constitutionality (in light of the same doctrine).51×51. See, e.g., Brief for Petitioner at 42–62, Bond v. United States, 134 S. Ct. 2077 (2014) (No. 12-158). At least in such cases, the record will be the same whichever argument the court ends up accepting, and that record will provide the court with the same information about the doctrine’s practical consequences.


After criticizing canons of constitutional avoidance for much of their article, Katyal and Schmidt take a surprising turn at the end: they offer a limited argument in favor of the canon about avoiding constitutional questions. They do not retreat from their view that the canon causes courts to “distort[]” statutes and to assert a mythical “rewriting power.”52×52. Katyal & Schmidt, supra note 1, at 2116. Ultimately, though, they endorse some distortion in the service of what academics call “underenforced constitutional norms.”53×53. Id. at 2159–60. The term comes from Lawrence Gene Sager, Fair Measure: The Legal Status of Underenforced Constitutional Norms, 91 Harv. L. Rev. 1212 (1978), although Dean Sager himself did not offer it as a reason for federal courts to interpret statutes to avoid constitutional questions. (The primary point of his article was not that federal courts should come up with alternative ways of enforcing “underenforced constitutional norms,” but rather that other governmental officials should not confuse “the legal scope of a constitutional norm” with “the scope of its federal judicial enforcement.” Id. at 1213; see id. at 1220–28.) The first article to enlist Sager’s concept in support of avoidance canons may have been Cass R. Sunstein, Interpreting Statutes in the Regulatory State, 103 Harv. L. Rev. 405, 468–69 (1989). In particular, Katyal and Schmidt adopt Professor Ernest Young’s argument that in some contexts, the canon about avoiding constitutional questions can be defended as a “resistance norm” that makes it harder for Congress to establish policies that the courts suspect might be unconstitutional but that the courts are not in a position to hold unconstitutional.54×54. See Young, supra note 15, at 1585–601.

I am not persuaded. As indicated above, I agree with Katyal and Schmidt that the canon about avoiding constitutional questions causes courts to distort statutes, and I do not think that Young’s argument justifies such distortion.

To be sure, I acknowledge that when courts are asked to apply a statute, they may not always be able to identify respects in which the statute conflicts with the Constitution; some constitutional norms will indeed go “underenforced” in judicial proceedings. Especially in the old days, that was partly because of concerns about the impropriety of accusing a legislature of having acted unconstitutionally.55×55. Until quite recently, for instance, courts were extremely reluctant to impute impermissible purposes to a duly enacted statute. When confronted with a statute that would be constitutional if enacted for one set of reasons but unconstitutional if enacted for a different set of reasons, courts often simply presumed that the legislature had behaved properly, without examining evidence to the contrary. See generally Caleb Nelson, Judicial Review of Legislative Purpose, 83 N.Y.U. L. Rev. 1784 (2008) (canvassing the history of this doctrine). Those concerns have faded,56×56. See id. at 1850–59. so there may now be fewer “underenforced constitutional norms” than there once were. But courts still face practical limitations on their ability to ferret out constitutional problems. Courts are neither omniscient nor omnicompetent, and the doctrinal tests that they develop may fail to detect some unconstitutionality.

It is also possible (though not obvious) that courts will have some sense of what they are missing. In particular, courts might know that certain categories of statutory provisions have higher-than-normal rates of undetected unconstitutionality. But it does not follow that courts should react by interpreting provisions in those categories more obtusely than usual.

Consider a hypothetical. Imagine that a court needs to interpret a statute that is not crystal clear. Under the ordinary techniques that courts use to identify the intended meaning of statutory language, Interpretation #1 seems best. But that reading of the statute would trench on an “underenforced constitutional norm” — meaning that although Interpretation #1 satisfies the doctrinal tests that courts use to implement the Constitution, there is some chance that it is nonetheless unconstitutional. What is more, an alternative reading is available: while the court thinks that members of the enacting legislature probably had Interpretation #1 in mind, they did not draft the statute clearly enough to rule out Interpretation #2. Should the court stick with Interpretation #1 (at the possible cost of applying a rule of decision that the legislature lacks the authority to establish), or should the court instead adopt Interpretation #2 (at the probable cost of enforcing something other than the statute’s intended meaning)?

If one assumes that Interpretation #2 is simply a narrower version of Interpretation #1, one might conceivably see Interpretation #2 as a good middle ground: Interpretation #2 might give the legislature much of what the legislature intended without risking an undetected violation of the Constitution.57×57. Cf. Philip P. Frickey, Getting from Joe to Gene (McCarthy): The Avoidance Canon, Legal Process Theory, and Narrowing Statutory Interpretation in the Early Warren Court, 93 Calif. L. Rev. 397, 450 (2005) (portraying the avoidance canon as “an alternative to both judicial review and routine statutory interpretation,” and explaining that “the canon reworks the statute by narrowing it — trimming away those aspects that threaten constitutional values while leaving the core of the statute in place”). But that rationalization will not always be available. What if Interpretation #2 is quite different than Interpretation #1 — so that the statute was intended to regulate spinach and courts are reading it to regulate peas?58×58. See supra note 33.

Even when Interpretation #2 is simply a subset of Interpretation #1, I am not sure why the existence of “underenforced constitutional norms” should lead courts to interpret the statute more narrowly than Congress probably intended (and than the courts would uphold). After all, under the best doctrinal tests that the courts have been able to develop, Interpretation #1 is constitutional. And although we are assuming that those tests yield a certain percentage of false negatives, one should not blithely tell courts to interpret all statutes in the relevant category stingily just to avoid the risk of missing some violations of the Constitution.

Indeed, that approach might itself raise constitutional concerns. As I understand the courts’ relationship to Congress, I am skeptical that courts should ever adopt an interpretive attitude of deliberate resistance to the directives that Congress probably intended to establish. If courts were to do so, I think that they would be “underenforcing” a different constitutional norm — the idea that courts are supposed to take directions from legislatures.

Professor Young likes the fact that the canon about avoiding constitutional questions “makes it harder — but still not impossible — for Congress to write statutes that intrude into areas of constitutional sensitivity.”59×59. Young, supra note 15, at 1552. It is certainly true that reading statutes to mean something other than what Congress probably intended makes it harder for Congress to accomplish what it wants. And when there is a nontrivial chance that what Congress wants would violate the Constitution in ways that courts would not detect, perhaps one can argue that Congress’s loss is our gain. But one should not assume that artificially narrow interpretations of statutes that trench on “underenforced constitutional norms” will always be better for society than giving effect to what Congress probably intended. While such resistance might sometimes block policies that would actually be unconstitutional, that result comes at the cost of blocking other policies too — policies that would be both constitutional and (in Congress’s judgment) good for society.

Given the complexity of modern statutory schemes, I suspect that the costs of “resistance norms” greatly exceed the benefits. At any rate, the benefits are speculative, and the costs are likely to be substantial. The courts’ departures from the intended meaning of statutory provisions would throw unpredictable wrenches into complicated systems, and sometimes would also have the arbitrary consequence of creating rules of decision for one set of cases when Congress had been trying to address a different set of cases. I would prefer courts simply to do the best job they can both in conducting judicial review and in interpreting statutes, without trying to compensate for their imperfections in one realm by magnifying their imperfections in the other.


*Emerson G. Spies Distinguished Professor of Law & Elizabeth D. and Richard A. Merrill Professor, University of Virginia School of Law. Thanks to John Harrison for helpful conversations.