OPINION EN BANC
PER CURIAM.In February 2002, the Rhode Island legislature adopted a redistricting plan in response to the 2000 census and a state constitutional amendment reducing the number of seats in both houses. Based on the allegations in the complaint, it appears that African-Americans are about 4 percent of Rhode Island’s population, but more than half live in Providence. Prior to redistricting, State Senate District 9 in Providence was 25.69 percent African-American and 41.08 percent Hispanic. Until redistricting, an African-American, Charles Walton, had represented District 9 for many years.
Under the 2002 redistricting plan, much of the same African American population now lies within the new District 2, which allegedly is 21.42 percent African-American and 46.74 percent Hispanic. In the 2002 primary after redistricting, a Latino challenger defeated Walton and went on to win the election. Well before the primary, in May 2002, a number of individual African-American voters and related organizations brought the present suit under section 2 of the Voting Rights Act, 42 U.S.C. § 1973 (2000), to challenge the redistricting plan.
In September 2002, the district court granted a motion under Fed.R.Civ.P. 12(b)(6) to dismiss the complaint, Metts v. Almond, 217 F.Supp.2d 252 (D.R.I.2002), holding that the claim failed two of the three threshold tests for a section 2 case under Thornburg v. Gingles, 478 U.S. 30, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). On appeal, a divided panel of this court reversed, remanding for further proceedings. Metts v. Murphy, 347 F.3d 346, 2003 WL 22434637 (1st Cir.2003).
*10We granted the defendants’ petition for rehearing en banc and vacated the panel opinion. Metts v. Murphy, No. 02-2204, 2003 U.S.App. LEXIS 24313 (1st Cir. Dec. 3, 2003). We now review and vacate the district court’s judgment of dismissal and remand for further proceedings. The reason for our remand is to allow a fuller development of the evidence, and further legal analysis based on that evidence, before any final determination is made.
Section 2, adopted as part of the Voting Rights Act of 1965, forbids voting-related measures that deny or abridge the right to vote “on account of race or color.” 42 U.S.C. § 1973. Under a 1982 amendment, a violation is established “if, based on the totality of circumstances, it is shown that ... members of a class of citizens ... have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” § 1973(b).
The Delphic language of the amendment can be understood only against the background of its legislative history and subsequent Supreme Court interpretation. The former tells us that discriminatory intent is not a necessary element in a violation and that Congress intended a broad range of factors to be taken into account. These points, and the relevant citations, are developed in Gingles, the first post-amendment decision on section 2 by the Court and still the leading authority. 478 U.S. at 43-46, 106 S.Ct. 2752.
However, Gingles was primarily concerned with the use of multi-member districts, which have an obvious potential to submerge the electoral power of even a substantial and cohesive minority bloc. 478 U.S. at 46-48 & nn. 11-13, 106 S.Ct. 2752. If such a group represents a majority of votes in a single member district but a numerical minority when combined with an adjoining district or districts, the combining of those districts into one multi-mem-ber district can easily eliminate the minority’s ability to elect one of their own to any of the seats.
In Gingles, the Supreme Court set up a three-part test, ruling that section 2 would ordinarily not be violated by multimember districts unless three conditions were met: that the minority challenging such a district would be “a majority” in a compact single member district; that the minority was politically cohesive (so it would bloc vote in such a district); and that the multi-district majority voted as a bloc (so it would usually defeat the minority’s candidate in a multi-member district). Gingles, 478 U.S. at 50-51, 106 S.Ct. 2752. If satisfied, these preconditions would not end the case but would raise a presumption of a violation. Vecinos De Barrio Uno v. City of Holyoke, 72 F.3d 973, 980 (1st Cir.1995); see also Johnson v. DeGrandy, 512 U.S. 997, 1011, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994) (Gingles preconditions necessary but not sufficient to establish claim).
Gingles was directed to a particular practice—multi-member districts—which the Court suggested was particularly problematic, 478 U.S. at 47-48, 106 S.Ct. 2752; Growe v. Emison, 507 U.S. 25, 40, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993), and the decision did not purport to offer a general or exclusive gloss on section 2 for all situations, Gingles, 478 U.S. at 46 n. 12, 106 S.Ct. 2752. But the concreteness of the Gingles test, set against the vagueness of the statute and plethora of criteria, has made it a focus in subsequent eases dealing with quite different problems. Indeed, the Supreme Court has said several times that Gingles applies to vote dilution claims directed against single member districts, see, e.g., Voinovich v. Quilter, 507 U.S. 146, 158, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993); Growe, 507 U.S. at 40-41, 113 S.Ct. *111075, but it has effectively qualified this statement in two different ways.
First, several Supreme Court opinions after Gingles have offered the prospect, or at least clearly reserved the possibility, that Gingles ’ first precondition — that a racial minority must be able to constitute a “majority” in a single-member district— could extend to a group that was a numerical minority but had predictable cross-over support from other groups. DeGrandy, 512 U.S. at 1008-09, 114 S.Ct. 2647; Voinovich, 507 U.S. at 158, 113 S.Ct. 1149 (“[T]he first Gingles precondition, the requirement that the group be sufficiently large to constitute a majority in a single district, would have to be modified or eliminated when analyzing the influence-dilution claim we assume, arguendo, to be actionable today.”). Further, the Court has so far reserved judgment on a second-cousin question: whether dilution of a minority racial group’s influence, as opposed to the power to elect, could violate section 2 — a position that would require substantial modification of Gingles ‘ first-prong “majority” precondition. Growe, 507 U.S. at 41 n. 5, 113 S.Ct. 1075; Vecinos De Barrio Uno, 72 F.3d at 990-91.
Second, where single member districts are at issue — as in our case — opinions have increasingly emphasized the open-ended, multi-factor inquiry that Congress intended for section 2 claims. Voinovich, 507 U.S. at 158, 113 S.Ct. 1149 (“Of course, the Gingles factors cannot be applied mechanically and without regard to the nature of the claim.”); DeGrandy, 512 U.S. at 1007, 114 S.Ct. 2647 (same). To say that Gin-gles applies as a precondition to section 2 liability may not tell one very much if Gingles itself is no longer to be “mechanically” applied. Gingles was in its original incarnation a mechanical first-step evaluation for a particular problem, so its rationale is not easily adapted by lower courts to a different set of problems.
The present case concerns not multi-member districts but a redrawing of single-member district boundaries. In one key district this has produced a modest readjustment in the proportionate sizes of the two large minority groups — but a readjustment that certainly can affect who wins the election. So far the parties’ argument has been about whether and how to squeeze this case into the Gingles preconditions — raising difficult questions about whether the “majority” requirement in Gingles is a numerical majority or an effective majority that could be constructed out of cross-over votes; how rigidly the Gingles preconditions apply when moving away from multi-member districts; and how to apply Gingles when no racial group makes up more than 50 percent of the district.
It is no accident that most cases under section 2 have been decided on summary judgment or after a verdict, and not on a motion to dismiss. This caution is especially apt where, as here, we are dealing with a major variant not addressed in Gingles itself — the single member district — and one with a relatively unusual history. As courts get more experience dealing with these cases and the rules firm up, it may be more feasible to dismiss weaker cases on the pleadings, but in the case before us we think that the plaintiffs are entitled to an. opportunity to develop evidence before the merits are resolved.
We are thus unwilling at the complaint stage to foreclose the possibility that a section 2 claim can ever be made out where the African-American population of a single member district is reduced in redistricting legislation from 26 to 21 percent. Yes, one would ordinarily expect the consequences to be small, but not always, and arguably not here (based on past his*12tory). At this point we know practically nothing about the motive for the change in district or the selection of the present configuration, the contours of the district chosen or the feasible alternatives, the impact of alternative districts on other minorities, or anything else that would help gauge how mechanically or flexibly the Gingles factors should be applied.
On the other hand, the plaintiffs cannot prevail merely by showing that an alternative plan gives them a greater opportunity to win the election, DeGrandy, 512 U.S. at 1017, 114 S.Ct. 2647 (“Failure to maximize cannot be the measure of § 2.”), or that an otherwise justified boundary change happened to cost African-Americans a seat. This would convert section 2’s all-circumstances test into the far more stringent “anti-retrogression” test of section 5, which imposes rigorous pre-clearance requirements on covered states to prevent redistricting plans with retrogressive consequences for African American voters. Compare 42 U.S.C. § 1973(a)-(b) (2000), with 42 U.S.C. § 1973c (2000). See generally Beer v. United States, 425 U.S. 130, 141, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976) (anti-retrogression test); Reno v. Bossier Parish Sch. Bd., 520 U.S. 471, 476-80, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997) (discussing differences between sections 2 and 5). Rhode Island is not a covered state.
As the district court correctly pointed out, there is tension in this case for plaintiffs in any effort to satisfy both the first and third prong of Gingles. To the extent that African-American voters have to rely on cross-over voting to prove they have the “ability to elect” a candidate of their choosing, their argument that the majority votes as a bloc against their preferred candidate is undercut. But it is not clear on the pleadings alone how many crossover votes are needed to win an election— unlike in Gingles, Rhode Island law allows a candidate to win with less than an absolute majority, see R.I. Const. art. IV, § 2 (general elections); R.I. Gen. Laws § 17-15-29 (2002) (primary elections) — nor do we have any evidence at this stage about how vigorously the majority votes as a bloc over time, nor the impact of the fact that the “majority” here is made up of both Hispanics and whites. Gingles itself warned that “there is no simple doctrinal test for the existence of legally significant racial bloc voting,” 478 U.S. at 58, 106 S.Ct. 2752, a further warning against deciding such issues in the abstract.
The burden of inquiry is on the plaintiffs — they are the ones challenging the redistricting plan — but in this case they are entitled (within ordinary limits) to develop the evidence that they think might help them. Whether a full-scale trial is needed is an entirely different matter; perhaps summary judgment will suffice depending on how the evidence develops and the ultimate theory or theories offered by both sides — theories that hopefully will go beyond dueling claims as to what Gingles means. In all events, it is premature to close the door now.
The judgment of the district court is vacated and the matter remanded for further proceedings consistent with this opinion. Each side shall bear its own costs on this appeal.
It is so ordered.