John Kilgour LENTELL, Brett Raynes and Juliet Raynes, Plaintiffs-Appellants, v. MERRILL LYNCH & CO. INC. and Henry M. Blodget, Defendants-Appellees, Thomas P. Willcutts, on behalf of himself and all others similarly situated, Yolanda Rice, individually and on behalf of all others similarly situated, Neil Trama, on behalf of himself and all others similarly situated, Brent Wickam, individually and on behalf of all others similarly situated, Marie Forte, on behalf of herself and all others similarly situated, C. Anthony Martignetti Trust, and on behalf of those similarly situated, Bob Raiano, individually and on behalf of those similarly situated, Christophe De Reynal, individually and on behalf of all others similarly situated, Diane Pilgrim, individually and on behalf of all others similarly situated, Turgut Er-gun, on behalf of himself and all others similarly situated, Doug Seiden-burg, individually and on behalf of all others similarly situated, Robert Rueben, on behalf of himself and all others similarly situated and Fulgham, individually and on behalf of all others similarly situated, Consolidated-Plaintiffs, Abraham Twersky Family Trust, on behalf of itself and all others similarly situated and John Deleo, on behalf of himself and all others similarly situated, Plaintiffs, 396 F.3d 161

JACOBS, Circuit Judge. John Kilgour Lentell and Brett and Juliet Raynes, as lead plaintiffs for purchasers of the publicly traded stock of two internet companies, appeal from the dismissal by the United States District Court for the Southern District of New York (Pollack, J.) of their securities-fraud actions against Merrill Lynch & Co. and its former star analyst, Henry M. Blodget (collectively, “Merrill Lynch,” “Merrill,” or “the Firm”).

In a nutshell, plaintiffs allege that Merrill, through Blodget and other research analysts, issued false and misleading reports recommending that investors purchase shares of 24/7 Real Media, Inc. (“24/7 Media”) and Interliant, Inc. (“Interliant”), even though the analysts did not then believe that those companies were a good investment. It is alleged that analysts were touted to investors as independent assessors of business prospects, but that they issued the falsely optimistic recommendations to cultivate the Firm’s investment-banking clients.

In a thorough opinion, Judge Pollack concluded: [i] that the suits were time-barred and (in any event) that they fail [ii] to plead loss causation, [iii] to plead fraud with the particularity required by Federal Rule of Civil Procedure 9(

b) and the Private Securities Litigation Reform Act of 1995 (“PSLRA”), and [iv] to overcome the “bespeaks caution” doctrine. We conclude that the underlying complaints were timely filed, but we affirm the dismissal on the ground that the complaints fail to plead that the alleged misrepresentations and omissions caused the claimed losses. BACKGROUND These securities-fraud suits arise from an investigation by the New York Attorney General (“NYAG”) into investment recommendations and research issued by prominent financial institutions, including Merrill Lynch.

The NYAG sought a state court order in April 2002 compelling the production of documents, testimony, and other evidence by Merrill Lynch and several of its current and former employees. The supporting affidavit outlined a scheme by Merrill Lynch’s research arm to publish bogus analysis in an effort to generate investment banking business. The NYAG’s papers cited dozens of internal communications that expressed bluntly negative views on internet stocks that the Firm’s analysts were then recommending to the investing public.

Within weeks, some 140 class-action complaints were filed, relying on the NYAG’s application to allege securities fraud in connection with Merrill Lynch’s analyses and investment recommendations concerning 27 publicly traded internet companies — including 24/7 Media and In-terliant. See In re Merrill Lynch & Co. Research Reports Sec. Litig., 273 F.Supp.2d 351, 357-59 (S.D.N.Y.2003).

The Judicial Panel on Multi-District Litigation (“MDL”) transferred these cases to Judge Pollack, see id., who consolidated the cases, appointed lead plaintiffs (by issuer), and ruled that the 24/7 Media and Interliant actions would proceed first and together. Id. at 359 n. 14. Amended, consolidated class-action complaints were filed in February 2003; the dispositive issue on appeal is the sufficiency of those complaints. I Because we assume plaintiffs factual allegations to be true on review of a motion to dismiss pursuant to Rule 12(b)(6), DeMuria v.

Hawkes, 328 F.3d 704, 706 (2d Cir.2003), the facts of Merrill Lynch’s fraud are taken from the amended complaints and any documents upon which they rely. See Rothman v. Gregor, 220 F.3d 81, 88-89 (2d Cir.2000). Merrill Lynch employs analysts to study and publish research and investment recommendations on a wide range of publicly traded companies. The Firm’s Internet Group covers so-called new economy companies that emerged in the 1990s as investment was ignited by electronic commerce and other internet-based business models.

Merrill Lynch is also an investment bank; among the services it provides in that capacity, Merrill assists companies seeking access to the capital markets by underwriting public offerings of securities. In theory, a “Chinese Wall” isolated Merrill’s Internet Group analysts from the investment bankers soliciting business from companies in the new economy. Plaintiffs claim that the Chinese Wall was breached. A.

The Alleged Fraud ■ Identical frauds are alleged as to 24/7 Media and Interliant: the publication by Merrill Lynch’s Internet Group of false and misleading research and investment recommendations “aimed at fraudulently driving up the market prices of [those] companies ... and motivated by the desire to obtain and maintain investment banking business for Merrill Lynch.” “The result of the scheme was to manipulate, inflate and maintain the market prices of the securities of the Internet companies at artificially high levels ... [and w]hen the market prices of the Internet companies fell, public investors lost hundreds of millions of dollars.” The complaints challenge approximately 80 reports issued during a combined class period of May 12, 1999 through February 20, 2001.

Merrill Lynch, 273 F.Supp.2d at 360. Henry Blodget — then a star analyst — headed the Internet Group throughout the putative class periods, and he figures prominently in plaintiffs’ allegations.

The scheme had five elements common to research published on 24/7 Media and Interliant: (i) “the public issuance and maintenance of knowingly or recklessly false, bullish research reports”; (ii) the publication of false “BUY or ACCUMULATE recommendations” on 24/7 Media and Interliant; (iii) the setting of “profoundly unrealistic price targets for [those] stocks”; (iv) the existence of undisclosed agreements between Merrill Lynch and 24/7 Media and Interliant to “ ‘trade’ favorable, bullish Analyst Reports for investment banking business directed to Merrill Lynch”; and (

v) the undisclosed “sharing of investment banking fees among Merrill Lynch and its internet analysts.” The false “buy” and “accumulate” recommendations appear in each of the challenged reports. Analyses issued on 24/7 Media and Interliant during the combined class periods were of three types: “Comments”; briefer, but largely similar “Bulletins”; and the terse “Morning Call Notes” (for 24/7 Media) and “Intra-Day Special Notes” (for Interliant).

Page one of every challenged Comment and Bulletin includes a four-barreled “Investment Opinion” expressed in the form “X-a-b-c” where (according to the margin notes) “X” is an “Investment Risk Rating” that ranged from “A” to “D”; “a” is a number keyed to intermediate “Appreciation Potential Rating,” ie., a prediction of the investment’s growth potential over the ensuing twelve months; “b” is a number keyed to long-term “Appreciation Potential Rating,” ie., a prediction of growth potential on a time-line greater than one year; and “c” is a number keyed to “Income Rating,” ie., a prediction of likely dividend payout.

Only the Appreciation Potential Ratings are alleged to have been false and misleading. Those ratings appeared in the full “Investment Opinion” offered in every challenged report, as well as in prominent, free-standing recommendations heading each Comment and Bulletin issued during the combined class period. According to the reports, appreciation potential was rated on a six-point scale: 1 — Buy; 2 — Accumulate; 3 — Neutral; 4 — Reduce; 5 — Sell; 6 — No Rating.

During the combined class period, the long-term and intermediate appreciation potentials for 24/7 Media and Interliant were never rated below “neutral,” and only rarely below “buy” or “accumulate.” Plaintiffs allege that this was de facto a 3-point ratings system, and that the ever-optimistic recommendations were bait and reward for investment-banking business. B. The 24/7 Media and Interliant Allegations 24/7 Media “provides marketing solutions to the digital advertising industry.” Merrill Lynch acted as lead underwriter for two public offerings made by 24/7 Media in August 1998 and April 1999.

Plaintiffs challenge as materially false and misleading each of the approximately 45 reports issued by Merrill’s Internet Group from May 12, 1999 through November 9, 2000. The stock-appreciation potential of 24/7 Media was rated at “accumulate” or “buy” throughout that period, until it was downgraded to “neutral” on November 9, 2000. The stock price gyrated from $45,125 on May 12, 1999, to a high of $64,625, and to a low of $2.9375 at the close of the putative class period.

According to plaintiffs, the 24/7 Media research reports — “particularly [the] ‘ACCUMULATE’ and ‘BUY’ recommendations” — were false and misleading, and failed to disclose that Merrill Lynch and Blodget “had a policy and practice throughout the Class Period of never issuing ... [a] rating or recommendation ... other than ‘BUY’ or ‘ACCUMULATE’” because to do so “would jeopardize Merrill Lynch’s ... ability to obtain underwriting or investment advisory engagements.” It is further alleged that the reports were issued primarily as a means to artificially inflate the price of 24/7 Media stock, and that the appreciation ratings were “nothing more than undisclosed ‘momentum’ plays — i.e. the stock should be bought because its price will rise, even though there are no rational economic reasons why the stock should trade at its current price ... [or] why the stock price should continue to rise.” Similar allegations are made with respect to the Internet Group’s coverage of Interliant, a provider of “enhanced Internet services that enable[d its] customers to deploy and manage their Web sites and network-based applications.” Merrill Lynch acted as co-lead underwriter of In-terliant’s initial public , offering in July 1999.

Plaintiffs challenge as false and misleading each of the approximately 35 reports issued by the Internet Group between August 4, 1999 (when coverage initiated with intermediate and long-term appreciation ratings of “Accumulate” and “Buy”) and February 20, 2001 (after which the stock was downgraded from “Buy/ Buy” to “Accumulate/Accumulate”). In-terliant was trading at $16,375 when Merrill initiated coverage, rose to a high of $55.50, and had plummeted to $4.00 as of February 21, 2001, the day after the putative class period closed.

Throughout this period, Merrill’s investment-banking arm assisted Interliant in its acquisition of 27 companies, and underwrote a $150 million convertible-bond offering in February 2000.

Plaintiffs challenge the veracity and completeness of the Interliant research reports in allegations virtually identical to those made regarding the 24/7 Media reports: the intermediate and long-term appreciation ratings were false and misleading; they were intended to artificially inflate Interliant’s share price and to encourage Interliant and other internet-sector companies to use Merrill Lynch for investment-banking services; and the ratings had no rational economic basis.

Plaintiffs filed amended, consolidated class-action complaints in February 2003 and Merrill promptly moved to dismiss for failure to state a claim. Judge Pollack granted defendants’ motion, citing numerous pleading deficiencies. This appeal followed. DISCUSSION We review de novo a district court’s dismissal of a complaint for failure to state a claim. Emergent Capital Inv. Mgmt., LLC v. Stonepath Group, Inc., 343 F.3d 189, 194 (2d Cir.2003).

The district court catalogued numerous deficiencies in the consolidated complaints, Merrill Lynch, 273 F.Supp.2d at 361-82; because we affirm their dismissal on the ground that plaintiffs failed to plead loss causation, we address only that issue and, antecedently, the statute of limitations. I “Section 10(

b) of the Securities Exchange Act of 1934 provides that ‘[n]o action shall be maintained to enforce any liability created under this section, unless brought within one year after the discovery of the facts constituting the violation and within three years after such violation.’ ” LC Capital Partners, LP v. Frontier Ins. Group, Inc., 318 F.3d 148, 154 (2d Cir.2003) (quoting 15 U.S.C. § 78i(e) (2000)).

The limitations period begins to run “after the plaintiff ‘obtains actual knowledge of the facts giving rise to the action or notice of the facts, which in the exercise of reasonable diligence, would have led to actual knowledge.’ ” Id. (quoting Kahn v. Kohlberg, Kravis, Roberts & Co., 970 F.2d 1030, 1042 (2d Cir.1992)) (emphasis omitted). Inquiry notice — often called “storm warnings” in the securities context — gives rise to a duty of inquiry “when the circumstances would suggest to an investor of ordinary intelligence the probability that she has been defrauded.” Levitt v.

Bear Steams & Co., Inc., 340 F.3d 94, 101 (2d Cir.2003) (quoting Dodds v. Cigna Sec., 12 F.3d 346, 350 (2d Cir.1993)); see also Newman v. Wamaco Group, Inc., 335 F.3d 187, 193 (2d Cir.2003) (“[T]he existence of fraud must be a probability, not a possibility.”).

In such circumstances, the imputation of knowledge will be timed in one of two ways: (i) “[i]f the investor makes no inquiry once the duty arises, knowledge will be imputed as of the date the duty arose”; and (ii) if some inquiry is made, “we will impute knowledge of what an investor in the exercise of reasonable diligence! ] should have discovered concerning the fraud, and in such cases the limitations period begins to run from the date such inquiry should have revealed the fraud.” LC Capital 318 F.3d at 154 (citation and internal quotation marks omitted).

Where inquiry notice is clearly established, see Newman, 335 F.3d at 193, dismissal of a securities-fraüd complaint as untimely may be readily affirmed; but “ ‘the applicable statute of limitations should not precipitate groundless or premature suits by requiring plaintiffs to file suit before they can discover with the exercise of reasonable diligence the necessary facts to support their claims,’ ” Rothman, 220 F.3d at 97 (quoting Sterlin v.

Biomune Sys., 154 F.3d 1191, 1202 (10th Cir.1998)). “[WJhether a plaintiff had Sufficient facts to place it on inquiry notice is ‘often inappropriate for resolution on a motion to dismiss.’ ” LC Capital, 318 F.3d at 156 (quoting Marks v. GDW Computer Ctr., Inc., 122 F.3d 363, 367 (7th-Cir.1997)).

In contrast, where “the facts needed for determination of when a reasonable investor of ordinary intelligence would have been aware of the existence of fraud can be gleaned from the complaint and papers ..-..integral to the complaint,” we can readily resolve the issue on a motion to dismiss, and have done so in “ ‘a vast number of cases.’ ” LC Capital, 318 F.3d at 156 (quoting Dodds, 12 F.3d at 352 n. 3).

The district court concluded that plaintiffs were on inquiry notice of Merrill’s alleged fraud “years prior to the filing of the[se] cases,” Merrill Lynch, 273 F.Supp.2d at 382, and dismissed the complaints as untimely filed. We disagree. Any fraud must be pled with particularity, Fed.R.Civ.P. 9(b); but the rule is applied assiduously to securities fraud. This Circuit’s strict pleading requirements in securities-fraud cases, see Novak v. Kasaks, 216 F.3d 300, 307-10 (2d Cir.2000), were (essentially) codified in the Private Securities Litigation Reform Act of 1995, id. at 309 — 11.

So no claim should be filed unless and until it can be supported by specific factual allegations. See, e.g., Levitt, 340 F.3d at 104 (“[C]omplaints in federal securities fraud cases [must] allege ‘those events which they assert give rise to a strong inference that [the] defendants had knowledge of th[e] facts ... or recklessly disregarded their existence,’ including ‘when the[ ] particular events occurred.’ ”) (quoting Ross v. A.H. Robins Co., 607 F.2d 545, 558 (2d Cir.1979)).

A ripe claim will keep only for one year, but “[t]he triggering ... data must be such that it relates directly to the misrepresentations and omissions the Plaintiffs allege in their action against the defendants.” Newman, 335 F.3d at 193 (emphasis added) (citation and quotation marks omitted). We have had frequent occasion to apply these rules. See, e.g., Levitt v. Bear Steams, Co., 340 F.3d 94 (2d Cir.2003); Newman v. Warnaco Group, Inc., 335 F.3d 187 (2d Cir.2003); LC Capital Part- nets, LP v. Frontier Ins. Group, Inc., 318 F.3d 148 (2d Cir.2003).

Our recent decisions reinforce the fact-specific nature of the limitations defense, particularly where the claim is foreclosed by inquiry notice. Storm warnings in the form of company-specific information probative of fraud will trigger a duty to investigate. For example, in LC Capital we concluded that three substantial charges taken against reserves by an issuer between 1994 and 1998 put plaintiffs on notice of probable wrongdoing more than a year before their untimely complaint was filed. LC Capital, 318 F.3d at 154-57.

Pleading with sufficient particularity may be especially difficult with claims against a “secondary” or “tertiary” wrongdoer (as opposed to an issuer or its officers or directors). See, e.g., Levitt, 340 F.3d at 102-04 (vacating the dismissal of a complaint against an issuer’s clearing agent where the district court failed to ascertain whether plaintiffs had access to facts sufficient to make out a claim of primary liability under § 10(b)).

We have been decidedly reluctant to foreclose such claims as untimely absent a manifest indication that plaintiffs “could have learned” the facts underpinning their allegations more than a year prior to filing. See id. No such clear indication appears in this record.

The fraud is alleged against a third party rather than against 24/7 Media or Interbank True, the Internet Group’s misleading statements and omissions were allegedly motivated by Merrill’s desire to win banking business from (inter alia) 24/7 Media and Interliant, but plaintiffs do not challenge any specific securities offering (or other investment-banking transactions) undertaken on behalf of either company. This is not a fraud that can be apprehended “simply by examining ... financial statements and media coverage” of the issuers. See Levitt, 340 F.3d at 103-04; cf.

LC Capital, 318 F.3d at 155 (probability of fraud could be gleaned from substantial reserves charges disclosed in issuer’s financial and other public statements). The 140 securities-fraud complaints consolidated before Judge Pollack were filed shortly after the NYAG sought to compel the production of documents and other evidence in its investigation of Merrill’s research practices. That investigation was undertaken pursuant to the Martin Act, N.Y. Gen Bus.

L. § 352 et seq., which “proscribes a wide array of business practices in connection with the sale of securities,” such as publication of fraudulent issuer-related research. The NYAG’s supporting affidavit catalogued many specific examples of such research issued by Merrill Lynch, not to prove a violation of federal securities law, but simply to compel additional discovery. Thus it was arguably sufficient for the NYAG to allege specific facts concerning Merrill’s coverage of one issuer to make a case for discovery pertaining to a wholly different issuer or issuers.

But such pleading does not suffice to plead federal securities fraud. The district court correctly consolidated the complaints issuer-by-issuer and required plaintiffs to allege facts “specific to the security in question,” including “who said what to whom concerning that particular security.” In re Merrill Lynch & Co., No. 02 MDL 1484, 2003 WL 253187, Case Management Order No. 3 (S.D.N.Y. Feb. 5, 2003) (emphasis added). By the same token, however, the one-year limitation period of § 10(

b) is triggered only by data that .“relates directly to the misrepresentations and omissions” that plaintiffs allege against Merrill Lynch. Newman, 335 F.3d at 193 (emphasis added) (citation and quotation marks omitted). The dispositive question is whether the data held sufficient by the district court meets this standard. Plaintiffs filed amended, consolidated class-action complaints in February 2003, which were dismissed as untimely four months later. Merrill Lynch, 273 F.Supp.2d at 382.

According to the district court, a duty to investigate the conflicts of interest among Merrill’s research analysts and investment bankers arose “years prior to the filing of the[se] cases” when numerous generic articles on the subject of structural conflicts appeared in the financial press. Id. The eleven articles cited by the court, published between May 2,1996 and June 12, 2000, were insufficient as a matter of law to put plaintiffs on inquiry notice of the frauds alleged with respect to the Internet Group’s coverage of 24/7 Media and Interliant. See id. at 382-89.

Many of the articles cited by the district court were published before 24/7 Media or Interliant went public. Pre-IPO articles could not prompt an investigation of the Internet Group’s coverage of 24/7 Media and Interliant because when the pieces appeared, plaintiffs could not have been holding the securities of either company, nor could Merrill Lynch have recommended them. The post-IPO articles are a closer question, as each describes (in a style echoed in the complaints) the conflicts of interest faced by a research analyst employed at a Wall Street investment bank.

For example, in October 1998 (two months after 24/7 Media’s IPO), Business Week reported that “the ‘Chinese Wall’ that on paper still separates a firm’s analysts from its investment bankers continues to crumble as analysts are encouraged to scout deals,” Merrill Lynch, 273 F.Supp.2d at 385 (quoting Jeffrey M. Ladderman, Who Can You Trust? Wall Street’s Spin Game, Bus.

Week, Oct. 5, 1998, at 148); that an observed consequence of this breakdown was the virtual elimination of “sell” ratings from the Wall Street analyst’s lexicon, see id. at 386-88; and that many banks (including Merrill) tied analysts’ compensation to the firm’s investment-banking income, id. In the district court’s view, this “plethora of public information would have required even a blind, deaf, or indifferent investor to take notice of the purported alleged ‘fraud,’ ” so that “[ejvery investor of reasonable intelligence would have been absolutely on inquiry notice.” Id. at 389 (emphasis in original).

Conflicts of interest present opportunities for fraud, but they do not, standing alone, evidence fraud — let alone furnish a basis sufficiently particular to support a fraud complaint. Nor does the existence of temptation trigger a duty of inquiry — at least, not by a reasonable investor. Something more than conflicted interest is required, no matter how well publicized the conflict may be.

Plaintiffs do allege something more: that Merrill’s analysts were actually corrupted as evidenced by investment opinions that were not just “systematically overly optimistic,” Merrill Lynch, 273 F.Supp.2d at 383 (quoting Steve Bailey & Steven Syre, Taking Analysts’ Tempting Forecasts with a Grain of Salt, Boston Globe, Oct. 23, 1996, at C1), but demonstrably false. In support, plaintiffs point to emails collected during the NYAG’s 2002 Martin Act investigation; the district court, however, found sufficient evidence of corruption in the public domain well before that investigation picked up steam.

The articles relied upon to support that finding fall well short of the specificity required to prompt further inquiry by a reasonable investor. The articles cited by the district court strongly suggest grounds to believe that certain investment recommendations were less than candid.

Well before the underlying complaints were filed, it was reported that “[ajnalysts routinely play up good news and sugarcoat the bad,” id. at 385 (citation omitted); that “[t]he analyst today is an investment banker in sheep’s clothing,” id. at 386 (citation omitted); that “[i]n public, Wall Street brokers say that their research is objective,” but “[pjrivately, they concede that ‘sell’ ratings are bad for investment-banking business,” id. (citation omitted); and that “too many analysts [are] keen to report that ‘what looks like a frog is really a prince,’ ” id. at 386-87 (citation omitted).

One anecdote goes beyond innuendo and metaphor: following Blodget’s decision to upgrade the investment recommendation on a particular stock, Blodget commented cheerily that “ ‘it’s dead money for a while, but I want to differentiate it from all the pieces of [expletive] we have buys on.’ ” Id. at 388 (quoting David Streitfeld, Analyst with a Knack for Shaking up Net Stocks; Henry Blodget is Wall Street’s Link Between Online Firms, Investors, Wash. Post, Apr. 2, 2000, at HI). However, that comment says nothing about 24/7 Media or Interli-ant; neither company is mentioned in any

article relied upon by the district court. If Blodget’s lone remark is sufficient to put a reasonable investor on inquiry notice of the frauds alleged in the 24/7 Media and Interliant complaints, then plaintiffs had a viable fraud claim with respect to every issuer covered by Merrill’s Internet Group no later than April 2, 2000. And if the conflicts of interest catalogued by the financial press were sufficient to trigger § 10(b)’s one-year limitation period, then the publication of a single investment recommendation by an underwriting bank would sustain a claim for securities fraud.

Such a result is incompatible with the congressional intent of the PSLRA “to deter strike suits wherein opportunistic private plaintiffs file securities fraud claims of dubious merit in order to exact large settlement recoveries.” Novak v. Kasaks, 216 F.3d 300, 306 (2d Cir.2000) (citation and quotation marks omitted).

We do not mean to suggest that inquiry notice could never be established on the basis of nonspecific public-pronouncements, but the level of particularity in pleading required by the PSLRA is such that inquiry notice can be established only where the triggering data “relates directly to the misrepresentations and omissions” alleged. Newman, 335 F.3d at 193 (emphasis added) (citation and quotation marks omitted); see also La Grasta v.

First Union Sec., Inc., 358 F.3d 840, 846 (11th Cir.2004) (finding earliest inquiry notice of stock analyst’s conflict of interest to be a published interview in which she referenced the conflict with respect to the specific security). The articles cited by the district court describe the conflicted situation of Wall Street’s research analysts; but evidence of the outright falsity of Merrill Lynch’s investment recommendations is stray and indiscriminate at best, and is insufficient to put plaintiffs on inquiry notice of the specific frauds alleged.

Furthermore, where (as here) plaintiffs’ allegations rely on internal communications that (arguably) could not be discovered absent a government-initiated investigation, we will not “punish [a] pleader for waiting until the appropriate factual information [has been] gathered by dismissing the complaint as time-barred.” Levitt, 340 F.3d at 104. For these reasons we reverse the district court’s ruling on the statute of limitations. We turn now to the sufficiency of plaintiffs’ timely allegations. II It is alleged (

i) that Merrill’s analysts did not actually believe 24/7 Media or In-terliant securities were a good investment when they encouraged the public to buy them; (ii) that the analysts’ reports failed to disclose that the Firm’s true motivation for publishing the fraudulent recommendations was to attract investment banking business; and (iii) that as a result of Merrill’s misstatements and omissions, plaintiffs bought the stocks and, when their value plummeted, lost millions of dollars. To state a claim for relief under § 10(

b) and Rule 10b-5, plaintiffs must allege that Merrill Lynch “(1) made misstatements or omissions of material fact; (2) with scienter; (3) in connection with the purchase or sale of securities; (4) upon which plaintiffs relied; and (5) -that plaintiffs’ reliance was the proximate cause of their injury.” In re IBM Securities Litigation, 163 F.3d 102, 106 (2d Cir.1998). The district court found the complaints deficient in numerous respects, including that plaintiffs failed to satisfy the particularity requirements of Rule 9(

b) and the PSLRA, or to overcome the “bespeaks caution” doctrine. Merrill Lynch, 273 F.Supp.2d at 368-78. We do not address these alternative bases for dismissal because, assuming away any other pleading defects, the district court correctly found that plaintiffs failed to plead that Merrill Lynch’s misstatements and omissions caused their investment losses. Id. at 362-68. It is long settled that a securities-fraud plaintiff “must prove both transaction and loss causation.” First Nationwide Bank v. Gelt Funding Corp., 27 F.3d 763, 769 (2d Cir.1994) (citing Citibank N.A. v.

K-H Corp., 968 F.2d 1489, 1495 (2d Cir.1992)); see also Mfrs. Hanover Trust Co. v. Drysdale Sec. Corp., 801 F.2d 13, 20-21 (2d Cir.1986); Schlick v. Penn-Dixie Cement Corp., 507 F.2d 374, 380 (2d Cir.1974). Transaction causation is akin to reliance, and requires only an allegation that “but for the claimed misrepresentations or omissions, the plaintiff would not have entered into the detrimental securities transaction.” Emergent Capital Inv. Mgmt., LLC v. Stonepath Group, Inc., 343 F.3d 189, 197 (2d Cir.2003).

Plaintiffs do not claim to have read Merrill’s reports, or to have bought 24/7 Media or Interliant securities through the Firm; instead, they rely on the fraud-on-the-market presumption blessed in Basic v. Levinson, 485 U.S. 224, 247, 108 S.Ct. 978, 99 L.Ed.2d 194 (1988). We assume for present purposes that the allegations could amount to a fraud on the market. Moreover, Merrill Lynch does not contest transaction causation at this stage, so the appellate issue is whether the complaints adequately plead loss causation.

Loss causation “is the causal link between the alleged misconduct and the economic harm ultimately suffered by the plaintiff.” Emergent Capital, 343 F.3d at 197. The PSLRA codified this judge-made requirement: “In any private action arising under this chapter, the plaintiff shall have the burden of proving that the act or omission of the defendant alleged to violate this

chapter caused the loss for which the plaintiff seeks to recover damages.” 15 U.S.C. § 78u-4(b)(4). We have described loss causation in terms of the tort-law concept of proximate cause, ie., “that the damages suffered by plaintiff must be a foreseeable consequence of any misrepresentation or material omission,” Emergent Capital, 343 F.3d at 197 (quoting Castella- no v. Young & Rubicam, 257 F.3d 171, 186 (2d Cir.2001)); but the tort analogy is imperfect.

A foreseeable injury at common law is one proximately caused by the defendant’s fault, but it cannot ordinarily be said that a drop in the value of a security is “caused” by the misstatements or omissions made about it, as opposed to the underlying circumstance that is concealed or misstated. Put another way, a misstatement or omission is the “proximate cause” of an investment loss if the risk that caused the loss was within the zone of risk concealed by the misrepresentations and omissions alleged by a disappointed investor. See AUSA Life Ins. Co. v.

Ernst & Young, 206 F.3d 202, 238 (2d Cir.2000) (Winter, J., dissenting). Thus to establish loss causation, “a plaintiff must allege ... that the subject of the fraudulent statement or omission was the cause of the actual loss suffered,” Suez Equity Investors, L.P. v. Toronto-Dominion Bank, 250 F.3d 87, 95 (2d Cir.2001) (emphasis added), ie., that the misstatement or omission concealed something from the market that, when disclosed, negatively affected the value of the security. Otherwise, the loss in question was not foreseeable.

We acknowledge that the pleading principles set out in the foregoing passage require both that the loss be foreseeable and that the loss be caused by the materialization of the concealed risk; and we further acknowledge that our opinion in Suez Equity can be (mis-)read to say that this Circuit has rejected the “materialization of risk” approach. Suez Equity does not purport to express this Circuit’s authoritative position, because that wording: (

i) is dicta consigned to a footnote; (ii) is framed in terms that are tentative and speculative, see id. at 98 n. 1 (“The standard that we have employed in this opinion attempts to reconcile what we view as our someivhat inconsistent precedents on loss causation.”) (emphasis added); and (iii) is expressly limited to what was (in 2001) “our precedents to date,” id. (emphasis added). This Court’s cases — post-Nwz and pre-Suez- — require both that the loss be foreseeable and that the loss be caused by the materialization of the concealed risk.

See Emergent Capital, 343 F.3d at 197 (“Similar to loss causation, the proximate cause element of common law fraud requires that plaintiff adequately allege a causal connection between defendants’ non-disclosures and the subsequent decline in ... value .... ”); id. at 198 (loss causation satisfied where the plaintiffs “specifically asserted a causal connection between the concealed information ... and the ultimate failure of the venture”); Castellano, 257 F.3d at 190 (“[a] jury could find that by failing to disclose material information ... [defendant] disguised the very risk to which [plaintiff] fell victim”); id. at 188 (“a jury could find that foreseeability links the omitted information and the ultimate injury in this case”); First Nationwide Bank, 27 F.3d at 769 (loss causation requires a showing “that the misstatements were the reason the transaction turned out to be a losing one”); Citibank, 968 F.2d at 1495 (“To establish loss causation a plaintiff must show, that the economic harm that it suffered occurred as a result of the alleged misrepresentations.”) (emphasis in original).

As this Court stated in Castellano: If the significance of the truth is such as to cause a reasonable investor to consider seriously a zone of risk that would be perceived as remote or highly unlikely by one believing the fraud, and the loss ultimately suffered is within that zone, then a misrepresentation or omission as to that information may be deemed a foreseeable or proximate cause of the loss. 257 F.3d at 188 (quoting AUSA Life Ins., 206 F.3d at 235 (Winter, J., dissenting)); see also Suez Equity, 250 F.3d at 97 (“it would have been foreseeable to defendants that facts concealed ... would have indicated [the executive’s] inability to run the Group, and would have forecast its (eventually fatal) liquidity problems”); id. at 98 (“Since defendants reasonably could have foreseen that [the executive’s] concealed lack of skill would cause the company’s eventual liquidity problems, defendants’ misrepresentations may be the causal precursor to the Group’s final failure.”).

But see, e.g., Marburg Management, Inc. v. Kohn, 629 F.2d 705, 708-10 (2d Cir.1980) (allegation that fraud induced investor to make an investment and to persevere with that investment sufficient to establish loss causation). We follow the holdings of Emergent Capital, Castellano, and Suez Equity.

Members of this Court have disagreed as to whether certain losses were attributable to a concealed risk, see AUSA Life Ins., 206 F.3d at 224-28 (Jacobs, J,, concurring in the mandate); but our precedents make clear that loss causation has to do with the relationship between the plaintiffs investment loss and the information misstated or concealed by the defendant. See Emergent Capital, 343 F.3d at 198-99; Castellano, 257 F.3d at 186-90; Suez Equity, 250 F.3d at 96-98.

If that relationship is sufficiently direct, loss causation is established, see, e.g., Suez Equity, 250 F.3d at 98 (finding that a CEO’s “concealed lack of managerial ability” induced the company’s failure); but if the connection is attenuated, or if the plaintiff fails to “demonstrate a causal connection between the content of the alleged misstatements or omissions and ‘the harm actually suffered,’ ” Emergent Capital, 343 F.3d at 199 (quoting Suez Equity, 250 F.3d at 96), a fraud claim will not lie.

See, e.g., Citibank, 968 F.2d at 1494-96 (finding defendant's nondisclosure of a seven-week bridge loan insufficiently connected to plaintiff's loss to establish causation). That is because the loss-causation requirement-as with the foreseeability limitation in tort-"is intended `to fix a legal limit on a person's responsibility, even for wrongful acts.'" Castellano, 257 F.3d at 186 (quoting First Nationwide Bank, 27 F.3d at 769-70). Loss causation is a fact-based inquiry and the degree of difficulty in pleading will be affected by circumstances, but our precedents establish certain parameters.

It is not enough to allege that a defendant’s misrepresentations and omissions induced a “purchase-time value disparity” between the price paid for a security and its “true ‘investment quality.’ ” Emergent Capital, 343 F.3d at 198 (clarifying Suez Equity, 250 F.3d at 97-99). Such an allegation-which is “nothing more than a paraphrased allegation of transaction causation”-explains why a particular investment was made, but does not speak to the relationship between the fraud and the loss of the investment. Emergent Capital, 343 F.3d at 198; see also Robbins v. Koger Props.

Inc., 116 F.3d 1441 (11th Cir.1997). “[I]f the loss was caused by an intervening event, like a general fall in the price of Internet stocks, the chain of causation ... is a matter of proof at trial and not to be decided on a Rule 12(b)(6) motion to dismiss.” Emergent Capital, 343 F.3d at 197.

However, “when the plaintiffs loss coincides with a marketwide phenomenon causing comparable losses to other investors, the prospect that the plaintiffs loss was caused by the fraud decreases,” and a plaintiffs claim fails when “it has not adequately ple[ ]d facts which, if proven, would show that its loss was caused by the alleged misstatements as opposed to intervening events.” First Nationwide Bank, 27 F.3d at 772.

Though all reasonable inferences are drawn in the plaintiffs favor on a motion to dismiss on the pleadings, “conclusions of law or unwarranted deductions of fact are not admitted.” Id. at 771 (citation omitted). Plaintiffs allege that when they invested, they were relying on the integrity of the market (including the fraudulent recommendations and omissions made by Merrill Lynch during the putative class periods), that the shares plummeted, and that their investments became virtually worthless.

To plead loss causation, the complaints must allege facts that support an inference that Merrill’s misstatements and omissions concealed the circumstances that bear upon the loss suffered such that plaintiffs would have been spared all or an ascertainable portion of that loss absent the fraud. As the district court found, no such allegations are made. Merrill Lynch, 273 F.Supp.2d at 367-68.

There is no allegation that the market reacted nega*tively to a corrective disclosure regarding the falsity of Merrill’s “buy” and “accumulate” recommendations and no allegation that Merrill misstated or omitted risks that did lead to the loss. This is fatal under Second Circuit precedent. As noted, to establish loss causation, “a plaintiff must allege ... that the subject of the fraudulent statement or omission was the cause of the actual loss suffered.” Suez Equity, 250 F.3d at 95 (emphasis added).

It is alleged that Merrill’s “buy” and “accumulate” recommendations were false and misleading with respect to 24/7 Media and Interliant, and that those recommendations artificially inflated the value of 24/7 Media and Interliant stock. However, plaintiffs do not allege that the subject of those false recommendations (that investors should buy or accumulate 24/7 Media and Interliant stock), or any corrective disclosure regarding the falsity of those recommendations, is the cause of the decline in stock value that plaintiffs claim as their loss.

Nor do plaintiffs allege that Merrill Lynch concealed or misstated any risks associated with an investment in 24/7 Media or Interliant, some of which presumably caused plaintiffs’ loss. Plaintiffs therefore failed to allege loss causation, as that requirement is set out in Emergent Capital, Castellano, and Suez Equity.

Plaintiffs do allege that Merrill’s “material misrepresentations and omissions induced á disparity between the transaction price and the true ‘investment quality’ ” of 24/7 Media and Interliant securities; “that the market price of [the] securities was artificially inflated”; and that the securities were acquired “at artificially inflated prices and [the plaintiffs] were damaged thereby.” Assuming (as we must) the truth of these allegations, they may establish transaction causation; but they do not provide the necessary causal link between Merrill’s fraud and plaintiffs’ losses. Emergent Capital, 343 F.3d at 198.

It is further alleged that plaintiffs were injured “because the risks that materialized were risks of which they were unaware as a result of Defendants’ scheme to defraud,” and that they would not have been injured absent the scheme. But that is a legal conclusion; missing are the necessary allegations of fact to support the conclusion. The only misrepresentation that can inhere to the “buy” and “accumulate” recommendations is that they were not Merrill’s true and sincere opinion.

Yet plaintiffs allege no loss resulting from the market’s realization that the opinions were false, or that Merrill concealed any risk that could plausibly (let alone foreseeably) have caused plaintiffs’ loss. In fact, as the district court recognized, plaintiffs fail to grapple in any meaningful way with the complexity of the reports that form the basis of their claims or, for that matter, to account for the price-volatility risk inherent in the stocks they chose to buy. See, e.g., Merrill Lynch, 273 F.Supp.2d at 367-68.

The essence of plaintiffs’ claim- is that the Internet Group’s ratings for medium— and long-term Appreciation Potential (ie., the “buy” and “accumulate” recommendations) issued during the putative class periods were false and misleading. Issue is taken with certain other -aspects of the reports, but plaintiffs do not challenge the detailed financial information and investment analysis published alongside Merrill’s fraudulent recommendations.

It is thus incontestible that the risk of price volatility — and hence, the risk of implosion — is apparent on the face of every report challenged in the underlying complaints. Merrill’s fraudulent “buy” and “accumulate” ratings appear as part of an “Investment Opinion” that includes an “Investment Risk Rating.” As described earlier, Merrill rates investment risk on a four-point scale, from “A” (“Low”), to “D” (“High”). 24/7 Media and Interliant were rated as D-grade investments, throughout the putative class periods. Merrill Lynch, 273 F.Supp.2d at 361.

Since the Investment Opinions are decoded in the margin of every “Bulletin” and “Comment” cited in the underlying complaints, the high-risk nature of the investment in 24/7 Media and Interliant was available to the marketplace just as readily as Merrill’s Appreciation Potential Ratings, along with all the other information contained in the challenged reports. See Basic, 485 U.S. at 247, 108 S.Ct. 978 (noting that “most publicly available information is reflected in market price”).

In addition to this systematic and consistent risk indicator, the research reports are full of (unchallenged) analysis, see Merrill Lynch, 273 F.Supp.2d at 367-68, suggesting that 24/7 Media and Interliant were volatile investments, and therefore subject to sudden and substantial devaluation risk. To plead successfully that Merrill’s fraud caused their losses, plaintiffs were required to allege facts to establish that the Firm’s misstatements and omissions concealed the price-volatility risk (or some other risk) that materialized and played some

part in diminishing the market value of 24/7 Media and Interbank We are told that Merrill’s “buy” and “accumulate” recommendations were false and misleading, and that the Firm failed to disclose conflicts of interest, salary arrangements, and collusive agreements among analysts, bankers, and 24/7 Media and Interbant. But plaintiffs nowhere explain how or to what extent those misrepresentations and omissions concealed the risk of a significant devaluation of 24/7 Media and Interbant securities.

The reports indicate that 24/7 Media and Interb-ant were high-risk investments, a designation that specifies, inter alia, a “high potential for price volatility,” and “no proven track record of earnings.” And the unchallenged financial analyses presented (fi.g., negative EPS ratios and consistent quarterly losses) certainly indicate weakness. Plaintiffs do not allege that Merrill “doctored” or hid, or omitted this information, all of which suggests that 24/7 Media and Interbant were volatile, devaluation-prone investments and that Merrill revealed as much in its reports.

This case is therefore sharply distinguishable from cases in which some or all of the risk that materialized was clearly concealed by a defendant’s misstatements or omissions. See, e.g., Suez Equity, 250 F.3d at 97-98; Emergent Capital, 343 F.3d at 196-98. We do not suggest that plaintiffs were required to allege the precise loss attributable to Merrill’s fraud, or that “systematically overly optimistic” ratings of the type published by the Internet Group are categorically beyond the reach of the securities laws.

But where (as here) substantial in-dicia of the risk that materialized are unambiguously apparent on the face of the disclosures alleged to conceal the very same risk, a plaintiff must allege (

i) facts sufficient to support an inference that it was defendant’s fraud — rather than other salient factors — that proximately caused plaintiffs loss; or (

n) facts sufficient to apportion the losses between the disclosed and concealed portions of the risk that ultimately destroyed an investment. Plaintiffs have done neither, and thus offer no factual basis to support the allegation that Merrill’s misrepresentations and omissions caused the losses flowing from the well-disclosed volatility of securities issued by 24/7 Media and Interbant. Finally, plaintiffs cast their claims in terms of market manipulation, pursuant to Rule 10b-5(

a) and (c). We hold that where the sole basis for such claims is alleged misrepresentations or omissions, plaintiffs have not made out a market manipulation claim under Rule 10b-5(

a) and (c), and remain subject to the heightened pleading requirements of the PSLRA. See Schnell v. Conseco, Inc., 43 F.Supp.2d 438, 447-48 (S.D.N.Y.1999) (B.D. Parker, /.) (refusing to characterize allegations as market manipulation claims where alleged “schemes to defraud” consisted largely of an aggregation of material misrepresentations to inflate stock, such as research reports containing misrepresentations of the underlying facts and use of false names to solicit investors). CONCLUSION For the foregoing reasons, the judgment is affirmed. .

For example, Merrill published a 24/7 Media "Bulletin'' on September 15, 1999, offering the Investment Opinion "D-l-1-9,” i.e., "High” investment risk, “Buy” for medium-term appreciation potential, "Buy” for long-term appreciation potential, and "No Cash Dividend” likely. The Appreciation Potential Ratings also appear as recommendations at the top right-hand corner on the front page, i.e., BUY and, immediately below, in smaller typeface, Long Term BUY. . Four articles appeared before 24/7 Media went public on August 13, 1998; seven went to press before Interliant’s public float on July 7, 1999. .

Defense counsel expressed skepticism at a similar line of argument; indeed, counsel was unwilling to concede any scenario in which complaints based on allegations of a type made by plaintiffs would be timely. We make no such categorical determination here. . Plaintiffs contend that they have alleged a corrective disclosure to the market, in alleging that Merrill’s eventual downgrades of 24/7 Media and Interliant stock (from "accumulate” to "neutral” and from "buy” to "accumulate,” respectively) negatively impacted the price of those securities.

These allegations do not amount to a corrective disclosure, however, because they do not reveal to the market the falsity of the prior recommendations. To the contrary, plaintiffs have argued (affirmatively) on this appeal that the falsity of Merrill’s recommendations was made public no earlier than April 2002, when the NYAG's affidavit ”describ[ed] the inner workings of Merrill’s Internet Group,” and that until then plaintiffs (and presumably the market at large) therefore lacked knowledge of the fraud. The complaints withstand the statute of limitations on the strength of that argument.

By the same token, however, Merrill’s concealed opinions regarding 24/7 Media and Interliant stock could not have caused a decrease in the value of those companies before the concealment was made public. . Plaintiffs challenge the reports' "bullish” price targets and the "Reason for Report” indicated on each, but do not challenge the substantive analysis offered with respect to 24/7 Media and Interliant. .

According to the Policy and Procedures Manual of Merrill Lynch’s Global Securities Research and Economics Group, a "D" Risk Rating indicates that a stock has "high potential for price volatility,” and that the issuer "may be unseasoned, [that it may] have a small [public] float, [that] management may be untested, [that] the industry may be new, [that] the company may depend heavily on one product or service, and/or [that the company] may not have a proven track record of earnings.” .' Report-specific indications of devaluation risk are abundant; for'example, in the 24/7 Media "Comment” and "Bulletin” issued August 12, 1999 (when the Intermediate Appreciation Potential Rating was raised from "Accumulate” to "Buy”), the final "Investment Highlight[]” featured on the opening page states “As always with this sector, the stock is likely to be extremely volatile.” The 24/7 Media reports consistently couch share-price targets in terms of the stock's valuation relative to competitors (a method plaintiffs do not challenge); and anticipated weaknesses (through, e.g., acquisitions) are reported throughout the putative class period.

The Interliant reports are peppered with similar analysis. Plaintiffs do not challenge or attempt to explain how this (and much additional) information bears upon their alleged losses.