dissenting.
The Court views this case as a dispute about the meaning of “subparagraph” in 15 U. S. C. § 1640(a)(2)(A). I think it involves more than that. For while I agree with the construction of that word adopted by the Court, see ante, at 60-62, by Justice Kennedy, see ante, at 66-67 (concurring opinion), and by Justice Thomas, see ante, at 67-68 (opinion concurring in judgment), I disagree with the conclusion that the Court believes follows. The ultimate question here is not the meaning of “subparagraph,” but the scope of the exception which contains that term. When is “liability under this subparagraph” limited by the $100/$1,000 brackets? In answering that question, I would give dispositive weight to the structure of § 1640(a)(2)(A), which indicates that the exception is part of clause (ii) and thus does not apply to clause (i).
After establishing the fact that “subparagraph” refers to a third-level subdivision within a section, denominated by a capital letter (here subparagraph (A)), see ante, at 60-62, the Court’s analysis proceeds in five steps. First, the Court presumes that this fact determines the scope of the exception. See ante, at 62. It does not. In context, the reference to “liability under this subparagraph” is indeterminate. Since it is not a freestanding limitation, but an exception to the liability imposed by clause (ii), it is quite possible to read it as saying that, in the consumer-lease cases covered by clause (ii), “the liability under this subparagraph” would be subject to the $100/$1,000 brackets. Using “subparagraph” in that way would hardly be nonsensical, since the only liability under subparagraph (A) that applies to consumer-lease cases is the amount of damages specified by clause (ii). In other words, if the exception is part of clause (ii), then “liability under this subparagraph” is actually synonymous with “liability under this clause,” cf. ibid., in the sense that either phrase would have the same effect were it to appear in clause *71(ii). As a result, the term “subparagraph” cannot end our inquiry.
The structure of subparagraph (A) provides the best indication of whether the exception is part of clause (ii). In simplified form, the subparagraph reads: “(i) . . . , (ii) ... , or (iii)_” Clauses (i), (ii), and (iii) are separated by commas, and an “or” appears before clause (iii). It is reasonable to conclude that the exception — which appears between “(ii)” and the comma that precedes “or (iii)” — is part of clause (ii). In fact, the Court admits in passing that the exception appears “in clause (ii).” Ibid, (emphasis added); see also ante, at 65 (Stevens, J., concurring) (referring to “the ceiling contained in (ii)” (emphasis added)). Yet the Court’s holding necessarily assumes that the exception somehow stands outside of clause (ii) — someplace where its reference to “subpar-agraph” can have a different effect than “clause” would. The Court effectively requires the exception to be either part of clauses (i) and (ii) simultaneously, or a part of subpar-agraph (A) that is not within any of the individual clauses. The legislative drafting manuals cited by the Court, see ante, at 60-61, and n. 4, reveal how unnatural such an unanchored subdivision would be. See L. Filson, The Legislative Drafter’s Desk Reference 223 (1992) (“If a section or other statutory unit contains subdivisions of any kind, it should never contain subdivisions of any other kind unless they are parts of one of those subdivisions” (emphasis added)); House Legislative Counsel’s Manual on Drafting Style, HLC No. 104-1, p. 24 (1995) (“If there is a subdivision of the text of a unit, there should not be a different kind of subdivision of that unit unless the latter is part of the 1st subdivision” (emphasis added)); Senate Office of the Legislative Counsel, Legislative Drafting Manual 10-11 (1997) (explaining how to avoid “using a cut-in followed by flush language,” that is, inserting a clause that is supposed to apply to (a)(1) and (a)(2) after (2) rather than between (a) and (a)(1)).
*72In its second step, the Court notes that, before 1995, the exception was generally read as applying to both clauses (i) and (ii). See ante, at 62. But the prior meaning is insufficient to reveal the meaning of the current version. As Justice Thomas points out, the placement of the exception “at the end of subdivision (A)” used to “indicate] that it was meant to refer to the whole of subdivision (A).” Ante, at 69. That inference, however, is no longer available, since Congress eliminated the “or” between clauses (i) and (ii) and added clause (iii). If the “or” were still there, it might just be possible to conceive of clauses (i) and (ii) as a sublist to which the exception attached as a whole. But one simply does not find a purportedly universal exception at the end of the second item in a three-item list.
The Court’s third step addresses clause (iii), which is not directly implicated by the facts of this case. The Court concludes that the underlying measure of damages in clause (i) (twice the finance charge) “continues to apply” to actions governed by the newly created clause (iii). Ante, at 62. That conclusion does not follow from merely reading the exception in clause (ii) to apply to clause (i), but it is necessary because, by reading “subparagraph” in the exception to have the effect of extending the exception to all of subparagraph (A), the Court has caused that exception to conflict with the higher limit in clause (iii). To remedy this, the Court proceeds (see ante, at 62-63, n. 9) to do further violence to § 1640(a)(2)(A), simply reading out its division into clauses (i), (ii), and (iii) entirely.1 It is not sound statutory construc*73tion to create a conflict by ignoring one feature of a statute and then to solve the problem by ignoring yet another. My construction of the exception in clause (ii) avoids the conflict altogether.
In its fourth step, the Court returns to the application of the $100/$1,000 brackets to clause (i). The Court finds “scant indication that Congress meant to alter the meaning of clause (i)” in 1995 and compares this to “ 'Sir Arthur Conan Doyle’s “dog that didn’t bark.’”” Ante, at 63 (quoting Church of Scientology of Cal. v. IRS, 484 U. S. 9, 17-18 (1987)). I hardly think it “scant indication” of intent to alter that Congress amended the text of the statute by moving the exception from the end of the list to the middle, making it impossible, without doing violence to the text, to read the exception as applying to the entire list. Needless to say, I also disagree with the Court’s reliance on things that the sponsors and floor managers of the 1995 amendment failed to say.2 I have often criticized the Court’s use of legislative history because it lends itself to a kind of ventriloquism. The Congressional Record or committeé reports are used to make words appear to come from Congress’s mouth which were spoken or written by others (individual Members of Congress, congressional aides, or even enterprising lobbyists). The Canon of Canine Silence that the Court invokes today introduces a reverse — and at least equally dangerous — phenomenon, under which courts may refuse to believe *74Congress’s own words unless they can see the lips of others moving in unison. See Morales v. Trans World Airlines, Inc., 504 U. S. 374, 385, n. 2 (1992) (“[Legislative history need not confirm the details of changes in the law effected by statutory language before we will interpret that language according to its natural meaning”).
In its fifth and final step, the Court asserts that it would be “anomalous” for liability to be “uncapped by the [$1,000] limit” when real property secures an open-end loan but capped by the $2,000 limit when it secures a closed-end loan, and that it would be “passing strange” for damages to be “substantially lower” under clause (iii) than under clause (i). Ante, at 63, and n. 10. The lack of a $1,000 limit does not, of course, make liability under clause (i) limitless. In all cases under clause (i), the damages are twice the finance charge, and the 1-year statute of limitations, 15 U. S. C. § 1640(e), naturally limits the amount of damages that can be sought.
More importantly, Congress would have expected the amounts financed (and thus the finance charges) under clause (i) to be generally much lower than those under clause (iii). In eases (like this one) where loans are not secured by real property, the amount financed can be no greater than $25,000. § 1603(3). Where loans are secured by real property, clause (iii) includes both first mortgages and second mortgages (or home equity loans), which are far more common and significantly larger than the open-end home equity lines of credit (HELOCs) that are still covered by clause (i). In 1994, 64% of home-owning households had first or second mortgages, but only 7% had HELOCs with outstanding balances. Survey Research Center, Univ. of Michigan, National Survey of Home Equity Loans 25 (Oct. 1998) (Table 1) (hereinafter National Survey). The mean first mortgage balance was $66,884; the mean second mortgage balance was $16,199; and the mean HELOC outstanding balance was *75$18,459. Ibid.3 Assuming a 10% interest rate (which would have been higher than a typical HELOC in 1994, see G. Can-ner & C. Luckett, Home Equity Lending: Evidence from Recent Surveys, 80 Fed. Res. Bull. 571, 582 (1994)), a year of finance charges on the mean HELOC would still have been less than $2,000 — which, when doubled, would still be less than two times the maximum damages under clause (iii), a disproportion no greater than what Congress has explicitly prescribed between clauses (ii) and (iii). In addition, very large outstanding balances on HELOCs are comparatively rare. In 2001, roughly 94% of them were less than the median outstanding mortgage principal of $69,227. See U. S. Census Bureau, American Housing Survey for the United States: 2001, pp. 150, 152 (Oct. 2002) (Table 3-15) (hereinafter American Housing Survey).4 Approximately 2% of HELOC balances were $100,000 or more (compared with approximately 32% of mortgages). See ibid. Because closed-end loans, are many times more common, and typically much larger, than open-end ones, the finance charges would generally be much higher under clause (iii) than under clause (i), providing a reason for Congress to focus more intently on limiting damages in clause (iii). As for the difference between clause (i) and the $1,000 cap in clause (ii): Consumer leases (principally car leases) are obviously a distinctive category and a special damages cap (which differs from clause (iii) as well as from clause (i)) no more demands an explana*76tion than does the fact that damages for those leases are tied to monthly payments rather than to finance charges. As Justice Stevens acknowledges, applying the $1,000 cap to clause (ii) but not clause (i) is a “plausible policy decision.” Ante, at 65. The Court should not fight the current structure of the statute merely to vindicate the suspicion that Congress actually made — but neglected to explain clearly— a different policy decision.
As the Court noted earlier this year: “If Congress enacted into law something different from what it intended, then it should amend the statute to conform it to its intent. It is beyond our province to rescue Congress from its drafting errors, and to provide for what we might think is the preferred result.” Lamie v. United States Trustee, 540 U. S. 526, 542 (2004) (internal quotation marks and alteration omitted). I would apply the exception only to the clause with which it is associated and affirm the judgment of the Court of Appeals.
In footnote 9, the Court asserts that its new reading merely requires one to pretend that “Congress had not added ‘(in)’ when it raised the cap on recovery.” That is not so — not, at least, if the Court adheres to the sound drafting principles that supposedly form the basis for its opinion. See supra, at 71. To adhere to those and also to apply both the limitation of clause (ii) and the limitation of clause (iii) to clause (i), one must “pretend” that .Congress not only had not added “(iii)” but also had eliminated “(i)” and “(ii).” Otherwise, those limits which are recited in clause (ii) would apply only to that clause.
The things that were said about the 1995 amendment are characteristically unhelpful. Rep. McCollum said: “[T]he bill raises the statutory damages for individual actions from $1,000 to $2,000.” 141 Cong. Rec. 26576 (1995); see also id., at 26898 (remarks of Sen. Mack) (same). Two weeks later, he “clarified]” his remarks by specifying that the amendments “apply solely to loans secured by real estate.” Id., at 27703 (statement of Reps. McCollum and Gonzalez). Taken literally, these floor statements could mean that the new $2,000 limit applies either to all “individual actions” under subparagraph (A), or to all “loans secured by real estate” under clauses (i) and (iii). Neither option is consistent with the Court’s conclusion that there is a $1,000 limit under clause (i).
The medians were, of course, lower than the means: $49,000 for first mortgages, $11,000 for second mortgages, and $15,000 for HELOCs. National Survey 25 (Table 1).
The 1994 survey did not report on the range of amounts owed on HELOCs. In 2001, however, the Census Bureau’s Housing Survey began reporting detailed data about HELOCs — in figures presumably comparable to the 1994 data recited above, since the median outstanding balance and median interest rate for HELOCs had not dramatically changed. (The 2001 medians were $17,517 and 8%. See American Housing Survey 152, 154 (Table 3-15).)