with whom Justice Souter joins, dissenting.
The Court today holds that 25 U. S. C. § 2719(d) (1994 ed.) clearly and unambiguously fails to give Indian Nations (Nations) the exemption from federal wagering excise and related occupational taxes enjoyed by the States. Because I believe § 2719(d) is subject to more than one interpretation, and because “statutes are to be construed liberally in favor of the Indians, with ambiguous provisions interpreted to their benefit,” Montana v. Blackfeet Tribe, 471 U. S. 759, 766 (1985), I respectfully dissent.
I
I agree with the Court that § 2719(d) incorporates an error in drafting. I disagree, however, that the section’s reference to chapter 35 is necessarily that error.
As originally proposed in the Senate, the bill that became the Indian Gaming Regulatory Act (IGRA) would have applied all gambling and wagering-related sections of the Internal Revenue Code to the Nations in the same manner as the States:
“Provisions of the Internal Revenue Code of 1986, concerning the taxation and the reporting and .withholding of taxes with respect to gambling or wagering operations shall apply to Indian gaming operations conducted pursuant to this Act the same as they apply to State operations.” S. 555, 100th Cong., 1st Sess., 37 (1987).
The Senate Indian Affairs Committee altered the language of this bill in two contradictory ways. It restricted the applicable Code sections to those relating to the “reporting and withholding of taxes with respect to the winnings” from gaming operations. 25 U. S. C. § 2719(d). It also added a parenthetical listing specific Code sections to be applied to the Nations in the same manner as the States, including *97chapter 35, a Code provision that relates to gambling operations generally, but not to the reporting and withholding of gambling winnings. Ibid.
One of these two changes must have been made in error. There is no reason to assume, however, that it must have been the latter. It is equally likely that Congress intended § 2719(d) to apply chapter 35 to the Nations, but adopted too restrictive a general characterization of the applicable sections.
The Court can do no more than speculate that the bill’s drafters included the parenthetical while the original restriction was in place and failed to remove it when that restriction was altered. See ante, at 92. Both the inclusion of the parenthetical and the alteration of the restriction occurred in the Senate committee, S. Rep. No. 100-446 (1988), and there is no way to determine the order in which they were adopted. If the parenthetical was added after the restriction, one could just as easily characterize the restriction as an unintentional holdover from a previous version of the bill.
True, reading the statute to grant the Nations the exemption requires the section’s reference to the “reporting and withholding of taxes with respect to the winnings” from gaming operations to sustain a meaning the words themselves cannot bear. But the Court’s reading of the statute fares no better: It requires excising from § 2719(d) Congress’ explicit reference to chapter 35. This goes beyond treating statutory language as mere surplusage. See Potter v. United States, 155 U. S. 438, 446 (1894) (the presence of statutory language “cannot be regarded as mere surplusage; it means something”); cf. ante, at 89. Surplusage is redundant statutory language, Babbitt v. Sweet Home Chapter, Communities for Great Ore., 515 U. S. 687, 697-698 (1995); W. Popkin, Materials on Legislation: Political Language and the Political Process 214 (3d ed. 2001) — the Court’s reading *98negates language that undeniably bears separate meaning. This is not a step to be undertaken lightly.
Both approaches therefore require rewriting the statute, see ante, at 89. Neither of these rewritings is necessarily more “serious” than the other: At most, each involves doing no more than reversing a change made in committee. Cf. ante, at 90.
The Court argues that, because the reference to chapter 35 occurs in a parenthetical, negating this language does less damage to the statute than concluding that the restrictive language outside the parenthetical is too narrowly drawn. I am aware of no generally accepted canon of statutory construction favoring language outside of parentheses to language within them, see, e.g., W. Eskridge, P. Frickey, & E. Garrett, Legislation and Statutory Interpretation, App. C (2000) (listing canons), nor do I think it wise for the Court to adopt one today. The importance of statutory language depends not on its punctuation, but on its meaning. See United States Nat. Bank of Ore. v. Independent Ins. Agents of America, Inc., 508 U. S. 439, 454 (1993) (“[A] purported plain-meaning analysis based only on punctuation is necessarily incomplete and runs the risk of distorting a statute’s true meaning”).
The fact that the parenthetical is illustrative does not change the analysis: If Congress’ illustration does not match its general description, there is as much reason to question the description as the illustration. Where another general description is possible — and was in fact part of the bill at an earlier stage — Congress’ choice of an example that matches the earlier description is at least ambiguous. Moreover, as §2719(d)’s parenthetical specifically lists statutory sections to be applied to the Nations, one might in fact conclude that the doctrine that the specific governs the general, Crawford Fitting Co. v. J. T. Gibbons, Inc., 482 U. S. 437, 445 (1987), makes this specific parenthetical even more significant than the general restriction that follows.
*99Nor is negating Congress’ clear reference to chapter 35 required by the policy behind the statute. If anything, congressional policy weighs in favor of the Nations. Congress’ central purpose in enacting IGRA was “to provide a statutory basis for the operation of gaming by Indian tribes as a means of promoting tribal economic development, self-sufficiency, and strong tribal governments.” § 2702(1). Exempting Nations from federal gaming taxation in the same manner as States preserves the Nations’ sovereignty and avoids giving state gaming a competitive advantage that would interfere with the Nations’ ability to raise revenue in this manner.
II
Because nothing in the text, legislative history, or underlying policies of § 2719(d) clearly resolves the contradiction inherent in the section, it is appropriate to turn to canons of statutory construction. The Nations urge the Court to rely upon the Indian canon that “statutes are to be construed liberally in favor of the Indians, with ambiguous provisions interpreted to their benefit,” Montana v. Blackfeet Tribe, 471 U. S., at 766, as a basis for deciding that the error in § 2719(d) lies in the restriction of the subclass, not in the specific listing of chapter 35. “[R]ooted in the unique trust relationship between the United States and the Indians,” County of Oneida v. Oneida Indian Nation of N. Y., 470 U. S. 226, 247 (1985), the Indian canon presumes congressional intent to assist its wards to overcome the disadvantages our country has placed upon them. Consistent with this purpose, the Indian canon applies to statutes as well as treaties: The form of the enactment does not change the presumption that Congress generally intends to benefit the Nations. Montana v. Blackfeet Tribe, supra; County of Yakima v. Confederated Tribes and Bands of Yakima Nation, 502 U. S. 251 (1992). In these cases, because Congress has chosen gaming as a means of enabling the Nations to achieve self-sufficiency, the Indian canon rightly dictates *100that Congress should be presumed to have intended the Nations to receive more, rather than less, revenue from this enterprise.
Of course, the Indian canon is not the only canon with potential applicability in these cases. Also relevant is the taxation principle, that exemptions from taxation must be clearly expressed. United States Trust Co. v. Helvering, 307 U. S. 57, 60 (1939); see also ante, at 95. These canons pull in opposite directions, the former favoring the Nations’ preferred reading, and the latter favoring the Government’s.
This Court has repeatedly held that, when these two canons conflict, the Indian canon predominates. In Choate v. Trapp, 224 U. S. 665 (1912), a State attempted to rely on the taxation principle to argue that a treaty provision making land granted to Indians nontaxable was merely a bounty, capable of being withdrawn at any time. The Court acknowledged the taxation principle, responding:
“But in the Government’s dealings with the Indians, the rule is exactly the contrary. The construction, instead of being strict, is liberal; doubtful expressions, instead of being resolved in favor of the United States, are to be resolved in favor of [Indian nations.]” Id., at 674-675.
In Squire v. Capoeman, 351 U. S. 1, 3 (1956), the Federal Government had conveyed land to the Nations “ ‘free of all charge or encumbrance whatsoever.’ ” Although this phrase did not expressly mention nontaxability, the Court held that the language “might well be sufficient to include taxation,” id., at 7. Invoking the Indian canon, id., at 6-7, we found the Nations exempt.
Likewise, in McClanahan v. Arizona Tax Comm’n, 411 U. S. 164 (1973), this Court inferred an exemption from state taxation of property inside reservations from a treaty reserving lands for the exclusive use and occupancy of the Nations. In doing so, the Court noted: “It is true, of course, that exemptions from tax laws should, as a general rule, be *101clearly expressed. But we have in the past construed language far more ambiguous than this as providing a tax exemption for Indians.” Id., at 176 (citing Squire, supra, at 100).
As the purpose behind the Indian canon is the same regardless of the form of enactment, supra, at 99, there is no reason to alter the Indian canon’s relative strength where a statute rather than a treaty is involved. Cf. ante, at 95. The primacy of the Indian canon over the taxation principle should not be surprising, as this Court has also held that the general presumption supporting the legality of executive action must yield to the Indian canon, a “counterpresumption specific” to Indians. Minnesota v. Mille Lacs Band of Chippewa Indians, 526 U. S. 172, 194, n. 5 (1999).
This Court has failed to apply the Indian canon to extend tax exemptions to the Nations only when nothing in the language of the underlying statute or treaty suggests the Nations should be exempted. The Cherokee Tobacco, 11 Wall. 616, 618, 620 (1871) (finding no exemption for the Nations from language imposing taxes on certain “ ‘articles produced anywhere within the exterior boundaries of the United States’ ”); Choteau v. Burnet, 283 U. S. 691, 693-694 (1931) (finding no exemption in provisions “subject[ing] the income of ‘every individual’ to tax,” including “income ‘from any source whatever’ ”); Superintendent of Five Civilized Tribes v. Commissioner, 295 U. S. 418 (1935) (same); Mescalero Apache Tribe v. Jones, 411 U. S. 145, 155 (1973) (refusing to exempt the Nations from taxes on land use income based on language that “[o]n its face . . . exempts land and rights in land, not income derived from its use”). Mescalero also went further, suggesting that because of the taxation principle, the Court would refuse to find such an exemption absent “clear statutory guidance.” Id., at 156. Mescalero’s formulation is admittedly in tension with the Court’s precedents giving the Indian canon primacy over the taxation principle where statutory language is ambiguous. As Mescalero was *102decided on the same day as one of those very precedents, the unanimous decision in McClanahan v. Arizona Tax Comm’n, supra, however, it cannot have intended to alter the Court’s established practice.
Section 2719(d) provides an even more persuasive case for application of the Indian canon than any of our precedents. Here, the Court is not being asked to create out of vague language a tax exemption not specifically provided for in the statute. Instead, the Nations simply ask the Court to use the Indian canon as a tiebreaker between two equally plausible (or, in these cases, equally implausible) constructions of a troubled statute, one which specifically makes chapter 35’s tax exemption applicable to the Nations, and one which specifically does not. Breaking interpretive ties is one of the least controversial uses of any canon of statutory construction. See Eskridge, Frickey, & Garrett, Legislation and Statutory Interpretation, at 341 (“The weakest kind of substantive canon operates merely as a tiebreaker at the end of the interpretive analysis”).
Faced with the . unhappy choice of determining which part of a flawed statutory section is in error, I would thus rely upon the long-established Indian canon of construction and adopt the reading most favorable to the Nations.