Judge Learned Hand observed that "it is one of the surest indexes of a mature and developed jurisprudence not to make a fortress out of the dictionary." The late respected Judge Hand notwithstanding, a fashion in statutory interpretation has arisen in the land, predicated on heavy reliance on the dictionary.

The "new textualists" hold, often harangue, that legal texts have a meaning that is to be discerned without the aid of any confusing legislative history, other extrinsic evidence of legislative intent, historical background, or most especially, consideration of the social consequences of one interpretation or another. The Supreme Court has chosen to make the Bankruptcy Code a kind of proving ground for "textualist" interpretation, regularly adopting textualist interpretations to settle the law on contested questions arising under the Bankruptcy Code.

Recently, with the generous financial support of the ABI Endowment Fund, I decided to test the quality of textualism as an interpretive method. The results are reported in "Textualism's Failures: A Study of Overruled Bankruptcy Decisions," 53 Vand. L. Rev. 887 (2000).

Like textualism, the pragmatic school of statutory interpretation, which has dominated American jurisprudence during most of the 20th century, generally focuses first on statutory text and context. Pragmatists, however, go further and consider sources outside the text in order to give concrete meaning to the statute in particular circumstances. Traditionally, these nontextual sources include legislative history, pre-enactment law and other circumstantial evidence of legislative intent examined with reference to the comparative policy implications of various interpretations.

The new textualists presume a lack of judicial competence to assess nontextualist sources and argue that textualism, despite its flaws, yields better results from a policy perspective than pragmatism. This essentially utilitarian defense of textualism is counterintuitive. On its face, pragmatism is more consistent with utilitarianism than textualism. If one accepts, as a utilitarian presumably does, that the goal of statutory interpretation is rational development of complex statutory schemes in a manner consistent with democratically selected policy goals, then all things being otherwise equal, pragmatic judges, adopting appropriately deferential attitudes toward political choices when informed by actual circumstances in real cases, should do better in achieving this goal than judges who defer to ex ante legislative drafting choices made without consciousness of the real consequences in the present unimagined or foggily anticipated case.

Determining error in the exegesis of a complex statutory scheme is generally a question of judgment and depends on an assessment of consequences and an understanding of the policies of the statute. Most interpretive "errors" are debatable.

Nevertheless, the textualist suggestion that legislative error correction is the appropriate safeguard to textualist excess suggests one reasonable, neutral (but partial) definition of error: interpretations subsequently superseded by statute. Statutory overruling indicates that the law as interpreted does not meet present political and policy goals in a substantial enough way to overcome the very considerable inertia within the legislative process. Given the subsequent legislation, the overruled interpretation is socially costly; the benefits of the overruled interpretation are not realized because of the later statutory change, but all the costs of adjudication and new legislation are paid. And there is an obvious injustice worked on the litigant who loses and others subject to the court decision that is destined for overruling, while future similarly situated persons obtain the benefit of the later amendment. If textualists make more errors that require statutory overruling, and if it is apparent that statutory overruling is costly, time-consuming and inefficient as error correction, these facts should be relevant to judges in choosing an appropriate method of statutory construction.

First, I identified and analyzed for interpretive method the 58 bankruptcy decisions under the 1978 Bankruptcy Code intentionally overruled by statute. For purposes of comparison, I then selected a group of random cases that Congress has not overruled from those federal appellate bankruptcy cases for which petitions for certiorari were acted upon by the Supreme Court between its 1982 and 1998 terms—a total of 804 cases. From this universe of cases, I drew six random samples of 30 cases each, for a total of 180 cases.

Comparing these two groups of cases—the overruled ones and the random ones—yields striking findings. Even a casual look at the data indicates that textualism is not a dominant mode of statutory interpretation in the courts of appeals, even in the bankruptcy area, notwithstanding the Supreme Court's particular insistence for a decade on "plain-meaning" constructions of the Bankruptcy Code. Approximately 80 percent of the control group cases were decided on largely or entirely nontextualist grounds.

On the other hand, of the 58 overruled cases, 31 (54 percent) were decided on textualist or primarily textualist grounds, while of the 52 cases in the control group, nine (18 percent) were decided on textualist or primarily textualist grounds. Three times as many overruled cases are textualist or primarily textualist. Conversely, 30 control group cases (58 percent) were decided on pragmatic or primarily pragmatic grounds. This is true of eight overruled cases (14 percent). That is, more than four times as many control group cases are pragmatic or primarily pragmatic.

Treating the categories of interpretive method as discrete and applying the chi-square test to determine statistical significance, these results are significant at the .001 level. Treating the categories of interpretive method as a continuum and applying the t-test procedure, the control group and overruled group exhibit approximately equal variances but significantly different means. The t-test confirms that the control group is significantly more pragmatic than the overruled group, with results at the .0001 level of significance.

There is virtually no chance that the larger number of pragmatic interpretations observed in the control group is a random event. The data strongly support the hypothesis that textualist decisions are more likely to be overruled by Congress than pragmatic ones.

Cases subsequently overruled by Congress presumably do not serve current policy goals well, at least in the view of Congress, which, after all, has the constitutional prerogative to set bankruptcy policy. The observation that overruled cases are disproportionately textualist in method should trouble utilitarian textualists. The observation is consistent with pragmatic intuitions that pragmatic interpretation leads to superior implementation of existing bankruptcy policy. Moreover, overruled cases are likely to be inconsistent with the intent of the enacting Congress as well. There is no evidence that as a general matter bankruptcy policy has sufficiently changed within the last 20 years such that, on average, honoring the intent of the drafters of the Bankruptcy Reform Act of 1978 would disserve current policy or vice-versa. Most of the overruled cases that form the subject of this research do not raise broad ideological issues, but rather deal with the details of bankruptcy administration and policy. Corrective legislation in such cases is likely to mean that the overruled case was inconsistent with the intent of the enacting Congress as well as current policy.

If one accepts the proposition that cases subsequently overruled by the Congress are not consistent with the legislative intent of the enacting Congress, the observation that overruled cases are disproportionately textualist in method should also trouble at least some judges who adopt textualism for legitimacy reasons. This observation is consistent with the pragmatic intuition that pragmatic interpretation leads on average to statutory interpretation that better approximates congressional intent.

The Bankruptcy Code is complicated and amendments often have unintended consequences, fail in part or sometimes fail completely. The median time required to overrule the 58 superseded cases was three years, with a minimum of less than one year and a maximum of 15 years. In 20 cases (34 percent), legislative correction took five or more years. Overruling by amending the Code is clearly a time-consuming business. Moreover, amending the Code is an uncertain and faulty business. The overrulings generated substantial further litigation or serious unintended consequences, or were botched, in 36 instances (62 percent). This appears to be equally true of amendments overruling textualist or primarily textualist decisions as it is of more pragmatic judicial decisions. Relying on legislation to correct judicial misinterpretations is like relying on surgery to correct ills caused by misprescribed drugs. It is costly, risky and rarely as efficacious as administering the correct medicine in the first place.

The evidence from bankruptcy shows that the clearest and most costly interpretive errors—the ones requiring legislative correction—are disproportionately textualist in origin. If we can agree that Congress and the courts should work together to achieve the most sensible and effective statutory scheme possible, the evidence is that over time, pragmatic interpretation appears more likely to lead to that result than textualism. Analysis of random bankruptcy cases indicates that pragmatic interpretive methods continue to be the norm in the bankruptcy field. Comparison of the random cases to the overruled ones indicates that pragmatism should remain the dominant mode of statutory construction.