Plaintiffs Claar, Mar, Young, Doll, Eggar, and Ellison appeal the district court’s grant of summary judgment for the defense in their action against their former employer, Burlington Northern Railroad, brought under the Federal Employees Liability Act (FELA), 45 U.S.C. § 51 et seq. The district court held inadmissible the expert testimony that plaintiffs offered in support of their claims, and granted summary judgment because, without that testimony, plaintiffs failed to show any causal connection between their workplace chemical exposure and their injuries. We affirm.
I
Plaintiffs are members of a group of twenty-seven people who brought suit against Burlington Northern, alleging that they suffer from a variety of ailments stemming from their exposure to a variety of chemicals while working at Burlington Northern’s Livingston, Montana, shop. Out of concern that plaintiffs might not be able to demonstrate a causal connection between their workplace chemical exposure and their injuries, the district court issued a ease management order consolidating the twenty-seven cases for pretrial purposes. The order required plaintiffs to submit affidavits describing their exposure to the chemicals they claim harmed them, and affidavits from physicians listing each plaintiffs specific injuries, the particular chemical(s) that in the physician’s opinion caused each injury, and the scientific basis for the physician’s conclusions. The court selected the cases of the present six plaintiffs for adjudication as test cases, and stayed the remaining cases.
Dr. Mark Hines prepared an affidavit concerning Eggar, and Dr. Richard Nelson prepared affidavits concerning the remaining five plaintiffs. The district court found these affidavits deficient because they failed to explain which chemical(s) caused which injuries, and failed to explain the scientific basis for the physicians’ conclusions that plaintiffs’ injuries were caused by workplace chemical exposure. Consequently, it ordered a second round of affidavits, again insisting that the physicians explain the basis for their opinions.
Approximately one month after receipt of the new affidavits, Burlington Northern moved for summary judgment on the ground that the affidavits were inadmissible. The district court agreed, finding that the affidavits still failed to explain Drs. Hines’s and Nelson’s reasoning and methods, and concluding that they lacked “the foundation and reliability necessary to support expert testimony.” The court held that, without the affidavits, plaintiffs could not demonstrate a causal relationship between chemical exposure and their injuries, and granted summary judgment for Burlington Northern. This appeal followed.
II
Plaintiffs first argue that the district court improperly relied on Fed.R.Evid. 703 for authority to scrutinize the reasoning and methodology underlying their affidavits. We review de novo the district court’s interpretation of the Federal Rules of Evidence, United States v. Williams, 989 F.2d 1061, 1070 (9th Cir.1993), and will uphold its decision to exclude expert testimony unless it is “mani*501festly erroneous,” Rogers v. Raymark Indus., Inc., 922 F.2d 1426, 1429 (9th Cir.1991).
Rule 703 provides, in relevant part:
The facts or data ... upon which an expert bases an opinion or inference may be those perceived by or made known to the expert at or before the hearing. If of a type reasonably relied upon by experts in the particular field in forming opinions or inferences upon the subject, the facts or data need not be admissible in evidence.
(Emphasis added.) To the extent that the district court relied on Rule 703, we agree that it was in error. The plain language of the rule permits no other conclusion. See Christophersen v. Allied-Signal Corp., 939 F.2d 1106, 1118 (5th Cir.1991) (Clark, C.J., concurring), cert. denied, — U.S. -, 112 S.Ct. 1280, 117 L.Ed.2d 506 (1992). Rule 703 merely relaxes, for experts, the requirement that witnesses have personal knowledge of the matter to which they testify. See Daubert v. Merrell Dow Pharmaceuticals, Inc., - U.S. -, -, 113 S.Ct. 2786, 2796, 125 L.Ed.2d 469 (1993); Fed.R.Evid. 602 (“A witness may not testify to a matter unless ... the witness has personal knowledge of the matter_ This rule is subject to the provisions of Rule 703, relating to opinion testimony by expert witnesses.”). It does not deal with the authority of the court to scrutinize an expert’s methodology or reasoning.
In evaluating the reliability of the affidavits offered by plaintiffs’ experts in this case, however, the district court relied not only on Rule 703, but also on Rule 702. And Rule 702 does provide authority for evaluating the reasoning and methodology underlying expert testimony.1 In fact, it requires a court to do so:
[T]he trial judge must ensure that any and all scientific testimony or evidence admitted is not only relevant, but reliable. The primary locus of this obligation is Rule 702.
Faced with a proffer of expert scientific testimony, then, the trial judge must determine at the outset, pursuant to Rule 104(a), whether the expert is proposing to testify to (1) scientific knowledge that (2) will assist the trier of fact to understand or determine a fact in issue. This entails a preliminary assessment of whether the reasoning or methodology underlying the testimony is scientifically valid and of whether that reasoning or methodology properly can be applied to the facts in issue.
Daubert, — U.S. at -, 113 S.Ct. at 2795-96 (emphasis added).2 Accordingly, the district court was both authorized and obligated to scrutinize carefully the reasoning and methodology underlying the affidavits offered by the plaintiffs in this case. The only question is whether it did so correctly.
Plaintiffs contend that the district court did not confine itself to examining their experts’ reasoning and methods, but instead based its decision to exclude the affidavits on improper assessments of the experts’ credibility. We disagree. It is true that, whenever a court rejects expert testimony because it is based on faulty methodology or reasoning, it follows implicitly that the expert’s conclusions are not to be credited. But as long as the court’s analysis focuses on the expert’s methods and reasoning, and not on the expert’s conclusions, its actions are proper. Daubert, — U.S. at -, 113 S.Ct. at 2797. Here, the district court confined itself to an inquiry into the reliability of the methods and reasoning underlying the conclusions in the proffered affidavits. Therefore, the court did not overstep its authority.
*502Fundamentally, the district court was concerned that the experts, Drs. Hines and Nelson, failed to explain the basis for their conclusions. .Before admitting the affidavits, the district court was affirmatively required to find that the experts’ conclusions were based on scientific knowledge. Id. at -, 113 S.Ct. at 2796. This requirement means that the court had to determine that Drs. Hines and Nelson arrived at their conclusions using scientific methods and procedures, and that those conclusions were not mere subjective beliefs or unsupported speculation. Id. at -, 113 S.Ct. at 2795. In an effort to make this determination, the district court repeatedly ordered the experts to explain the reasoning and methods underlying their conclusions. Despite those orders, the affidavits are devoid of any such explanation. Consequently, the district court could not make the findings required by Rule 702, and therefore properly refused to admit the affidavits into evidence. We find no error in the district court’s decision.
The district court made other findings that also focus exclusively on the experts’ methods and that support its decision to exclude the affidavits. For example, it found that neither Dr. Hines nor Dr. Nelson made any effort to rule out other possible causes for the injuries plaintiffs complain of, even though they admitted that this step would be standard procedure before arriving at a diagnosis.3 See Daubert, — U.S. at -, 113 S.Ct. at 2797 (use of generally accepted scientific methods one indication of reliability). For example, Dr. Nelson diagnosed Eggar as having “dyscalculia” (poor arithmetic ability) and “spelling dispraxia” (poor spelling ability), without first reviewing Eggar’s school records, which indicate that he suffered from these “conditions” since childhood. Because he failed to review these records, Dr. Nelson could testify reliably only that Eggar’s spelling and arithmetic abilities are below average. He offered no scientific basis for concluding that chemical exposure played any part at all in Eggar’s condition.4
Similarly, the district court found that the toxicology sections of the affidavits, which summarize the scientific literature discussing evidence that particular chemicals can cause particular injuries (and which thus are a necessary foundation for any conclusions the experts might draw), fail to discuss the majority of the medical conditions alleged by plaintiffs. Consequently, the court correctly determined that there is no evidence that the experts’ conclusions about the cause of these conditions are based on anything more than subjective belief and unsupported speculation.
The district court also found that Drs. Hines and Nelson formed their opinions before reading the relevant literature, even though they admitted that they were not sufficiently familiar with the field to diagnose the causes of plaintiffs’ injuries without first reviewing that literature.5 The district court correctly considered this finding, which sheds light on the experts’ methods, in making its admissibility determination. “In order to qualify as ‘scientific knowledge’ [as required by Rule 702], an inference must be derived by the scientific method.” Daubert, — U.S. at -, 113 S.Ct. at 2795. Coming to a firm conclusion first and then doing research to *503support it is the antithesis of this method. See Viterbo v. Dow Chemical Co., 646 F.Supp. 1420, 1425 (E.D.Tex.1986), aff'd, 826 F.2d 420 (5th Cir.1987). Certainly, scientists may form initial tentative hypotheses. However, scientists whose conviction about the ultimate conclusion of their research is so firm that they are willing to aver under oath that it is correct prior to performing the necessary validating tests could properly be viewed by the district court as lacking the objectivity that is the hallmark of the scientific method.
We conclude that the district court correctly applied Rule 702, and did not manifestly err in rejecting the proffered expert testimony as unreliable.6
Ill
Plaintiffs next argue that, even if the district court otherwise correctly applied Rule 702, it erred in demanding that the experts demonstrate a causal connection between specific injuries and specific chemicals, contending that FELA “requires no such specific causation evidence.”7 This argument misconceives both the standard of causation under FELA and its relationship to the Federal Rules of Evidence.
The standard of causation under FELA and the standards for admission of expert testimony under the Federal Rules of Evidence are distinct issues and do not affect one another. Cf. Fed.R.Evid. 1101(b) (rules apply equally to both civil and criminal proceedings even though the burdens of proof are different). It is true that under FELA the quantum of evidence sufficient to present a jury question of causation is less than it is in a common law tort action. Pierce v. Southern Pacific Trans. Co., 823 F.2d 1366, 1370 (9th Cir.1987). Under FELA, the jury should determine liability so long as the evidence justifies “with reason, the conclusion that employer negligence played any part, even the slightest, in producing the injury.” Rogers v. Missouri Pacific R.R. Co., 352 U.S. 500, 507, 77 S.Ct. 443, 449, 1 L.Ed.2d 493 (1957). This does not mean, however, that FELA plaintiffs need make no showing of causation. Nor does it mean that in FELA eases courts must allow expert testimony that in other contexts would be inadmissible. It means only that in FELA eases the negligence of the defendant “need not be the sole cause or whole cause” of the plaintiffs injuries. See Oglesby v. Southern Pacific Trans. Co., 6 F.3d 603, 608 (9th Cir.1993). FELA plaintiffs still must demonstrate some causal connection between a defendant’s negligence and their injuries. Id.; Mayhew v. Bell S.S. Co., 917 F.2d 961, 964 (6th Cir.1990);8 Edmonds v. Illinois Central Gulf R.R., 910 F.2d 1284, 1288 (5th Cir.1990); Moody v. Maine Central R.R. Co., 823 F.2d 693, 695 (1st Cir.1987).9
Gallick v. Baltimore & O.R. Co., 372 U.S. 108, 83 S.Ct. 659, 9 L.Ed.2d 618 (1963), and Lavender v. Kurn, 327 U.S. 645, 66 S.Ct. 740, *50490 L.Ed. 916 (1946), do not compel a different conclusion. These cases arguably stand for the proposition that, in order to create an issue for the jury, a FELA plaintiff need show only the mere possibility that a defendant’s actions caused his injuries. Plaintiffs here contend that they satisfied this minimal standard by establishing that they were exposed to some chemicals that can cause some of the injuries they complain of, and that the district court erred in demanding that their experts further demonstrate a causal connection before letting the case go to the jury.
Gallick and Lavender, however, involved situations in which no special expertise was necessary to draw a causal inference. Once evidence was presented that the plaintiffs injuries just possibly may have been caused by the railroad’s negligence, the jury in those eases was qualified to determine whether the railroad’s negligence in fact played a part in causing the injuries. Here, in contrast, expert testimony is necessary to establish even that small quantum of causation required by FELA. See Moody, 823 F.2d at 695 (if drawing a particular conclusion requires specialized knowledge, expert testimony is required) (citing W.P. Prosser and Keeton on Torts 269 (5th ed. 1984)); Mayhew, 917 F.2d at 963. Plaintiffs, not having proffered any admissible expert testimony, have no evidence that workplace exposure to chemicals played any part, no matter how small, in causing their injuries. Accordingly, there is no genuine issue of material fact concerning even the modicum of causation FELA requires, and summary judgment was appropriate.
IV
Finally, plaintiffs contend that at least portions of their experts’ affidavits are admissible. They argue in effect that the district court threw out the baby with the bath water, and insist that the district court was “required carefully to consider each claim of each plaintiff to determine the admissibility of expert testimony regarding that claim.” Yet they point to no place in the record where a particular claim of a particular plaintiff is supported by admissible expert testimony. Instead, they merely summarize some of the exposure descriptions contained in the affidavits, some of the toxicity studies (which show that some of the chemicals to which some of the plaintiffs were exposed can cause some of the injuries complained of), and then cite Dr. Nelson’s and Dr. Hines’s inadmissible conelusory opinions that workplace chemical exposure caused their injuries. Because Burlington Northern pointed out the lack of admissible evidence on the issue of causation, plaintiffs had the burden of advertising to “specific facts showing that there is a genuine issue for trial.” Celotex Corp. v. Catrett, 477 U.S. 317, 324, 106 S.Ct. 2548, 2553, 91 L.Ed.2d 265 (1986); Fed. R.Civ.Proc. 56(e). Plaintiffs failed to carry this burden. It is not the district court’s job to sift through the record to find admissible evidence in support of a non-moving party’s case.
Y
We deny plaintiffs’ request that we give them still another chance to submit admissible affidavits. Parties are not permitted to file late affidavits in support of their opposition to a motion for summary judgment without invoking Fed.R.Civ.Proc. 56(f) and indicating why they cannot timely file the required affidavits. Carpenter v. Universal Star Shipping, S.A., 924 F.2d 1539, 1547 (9th Cir.1991), cert. denied, - U.S. -, 113 S.Ct. 413, 121 L.Ed.2d 337 (1992). Plaintiffs did not invoke Rule 56(f), nor have they suggested any reason for their failure to file timely, acceptable affidavits. They had two opportunities to produce affidavits, and were specifically instructed as to what these must contain. They had the additional opportunity, but did not avail themselves of it, to move for leave to produce stronger affidavits after Burlington Northern filed its motion for summary judgment. We decline to permit them to remedy their deficiency at this late stage of the proceedings.
CONCLUSION
To survive Burlington Northern’s motion for summary judgment, plaintiffs were required to offer admissible expert testimony showing that exposure to chemicals in the workplace played some part in producing *505their injuries. The district court properly-scrutinized the reasoning and methodology underlying the expert testimony proffered by plaintiffs, and did not manifestly err in holding it inadmissible.
AFFIRMED.