concurring.
For the first time since judges were admonished to guard the gate leading to the expert witness stand, this Court must review whether some evidence connects a particular chemical exposure to a particular disease. In Daubert v. Merrell Dow Pharmaceuticals, Inc.1 and Merrell Dow Pharmaceuticals, Inc. v. Havner,2 our highest courts set out in detail the standards we must apply. But the Court balks, choosing instead to follow its own brand-new standards. As an intermediate appellate court, this is not our role. Thus, I write separately.
1. What Daubert and Havner Require
Every court superior to ours (and there are at least three) has adopted a specific list of factors that courts “should consider” when evaluating scientific medical testimony.3 The Court considers these factors only long enough to reject them. It is trae these factors may not apply in all cases, *613but we can hardly ignore them when — -just as in Daubert and Havner — scientific testimony attempts to link a chemical exposure to a medical condition.
Dr. David Miller was Anderson’s sole expert witness on causation. He testified that Anderson’s lung disease — bronchioli-tis obliterans organizing pneumonia (known by its unfortunate acronym “BOOP”) — was caused by naphtha, even though he did not know the chemicals in naphtha, the level of Anderson’s exposure, or any medical literature that connected the two. He stated two reasons.
First, he relied on the temporal connection between Anderson’s exposure and his disease. Such post-hoc-ergo-propter-hoc4 reasoning is sometimes correct, sometimes fallacious. Proximity in time may suggest proximate cause when the harm from exposure is immediate, direct, and obvious, such as coughing, watery eyes, or (in Anderson’s case) headaches and nausea. But when the alleged harm is cancer, a birth defect, or BOOP — conditions whose development may be delayed, indirect, or largely unknown — something more is required.5
Second, he used a “differential diagnosis,” based partly on medical tests, but mostly on what Anderson told him. The Court finds this “differential diagnosis” reliable. For several reasons, I am not so sure.
First, this was no diagnosis. Differential diagnosis is a “determination of which of two or more diseases with similar symptoms is the one from which the patient is suffering.”6 But there is only one way to diagnose BOOP — a biopsy.7 Dr. Miller’s process of elimination led him to suspect a generalized pneumonia, but only the biopsy diagnosed Anderson’s disease as BOOP. Maybe courts should refuse to scrutinize differential diagnoses when they are used “to prescribe medical treatment with potential life-or-death consequences.” But that’s not what happened here.8
Second, Dr. Miller’s opinion was critical for something else — Anderson’s million-dollar claim against his employer.9 To the extent his opinion relied on objective medical tests, this is fair game — let Coastal disprove these if it can. But the opinion *614relied heavily on Anderson’s truthfulness.10 Doctors are no experts at that.11
Third, no evidence shows that Dr. Miller’s reliance upon temporal proximity and a process of elimination is a rehable method used by scientists to establish what causes BOOP, especially when a new “cause” is involved. This is demonstrated by applying the factors Daubert and Hav-ner say we “should consider”:
• Testing. One could test whether this methodology is a valid way to find new causes of BOOP, or whether naphtha causes BOOP, but apparently no one has done so;
• Peer review and publication. Neither Dr. Miller’s methodology nor conclusion have ever been published or reviewed by his peers;12
• Potential rate of error. For the above reasons, this is unknown;
• General acceptance. Without citation to any medical studies, there appears to be no general acceptance of this testimony from other members of the medical community;13 and
• Non-judicial use. As previously mentioned, the primary use of this testimony was Anderson’s lawsuit — it had no effect on his treatment.
Moreover, the peculiar facts here make a differential diagnosis particularly subjective and unreliable. As noted in Havner, there is a relationship between mathematical percentages and the burden of proof.14 Dr. Miller admitted that the cause in about a third of all BOOP cases was unknown, exceeding the incidence of any other known cause. Thus, even if Anderson had one of the conditions known to be associated with BOOP (say, a viral infection or arthritis), without knowing more it is hard to see how he could prove this cause was “more probable” than all the other unknown causes. A probability smaller than one-third simply cannot also be more probable than one-third.
Finally, in this case there are both unknown and ubiquitous causes, both of which it is impossible to “rule out.” For the former, one cannot develop a checklist of symptoms, exposures, or medical history to exclude them. For the latter, one cannot exclude something that everyone is *615exposed to. In this case, Dr. Miller never ruled out nitrogen oxides (admitting they were everywhere), or the unknown causes that predominate in BOOP cases; he simply ignored them. This is not proper methodology.15
Differential diagnoses may sometimes meet the Daubert standards. If the percentage of unknown causes is small, a known cause may well be “more likely.” 16 If there are only four causes of a disease, ruling out three may tell us all we need to know. But ruling out three proves little if the number of causes is twenty, or unknown. Because Dr. Miller’s elimination of some causes in Anderson’s case cannot tell us (in reasonable probability) the actual cause, it is no evidence of causation, specific or otherwise.17
II. The Court’s New Differential-Diagnosis Rule
Instead of testing Dr. Miller’s “differential diagnosis” by the factors Daubert and Havner say we should consider, the Court holds that no differential diagnosis need comply with them. This stretches “flexible” too far.18 Not all fields of expertise follow scientific principles or have peer-reviewed journals, but that is no license to ignore these factors in fields that do.
The Court rejects the Daubert and Hav-ner factors because they are “hard science,” opting instead for a “soft science” approach to differential diagnoses. I may be mistaken, but I believe most scientists that fly on airplanes, go to the doctor, or get sued for a million dollars would expect the science to be pretty “hard.” True, doctors in a clinical setting may sometimes have to guess. Exigencies may require them to try certain treatments before they know what will work. But the Court does not explain why these lower standards make them reliable in court.
Instead, the Court declares a new test— differential diagnoses are admissible if “properly conducted and explained” or “properly relie[d] upon and/or utilize[d].” This is not much of a test. At the least, it is a very subjective one — who is to say what is proper? The Daubert and Havner tests contain objective standards to ensure reliability; the Court’s test returns us to the days of “flexible” science.19
There is (as the Court notes) considerable disagreement about whether and to what extent differential diagnoses meet the requirements of Daubert and Havner. In part, this is because they come in many varieties and circumstances. For example, *616excluding potential causes by laboratory tests is surely more reliable than excluding them based on subjective observations. Relying on a patient’s reported history may be reliable when the risk factors carry no moral or legal baggage with them, but less reliable when they do. And the context of the diagnosis (whether for treatment, for insurance coverage, or for litigation) may also affect reliability.20 Given the unsettled state of the law, it would be wiser to decide this case on its facts, rather than rushing to the universal conclusion that differential diagnoses establish specific but not general causation.21
The Court’s commitment to and analysis of differential diagnoses is impressive, especially as very little of it comes from the parties’ briefs.22 “No one doubts the utility of medical histories in general or the process by which doctors rule out some known causes of disease in order to finalize a diagnosis.”23 But the question is not whether doctors in general ought to use differential diagnoses; the question is whether scientists recognize differential diagnosis as a proper method for determining a new cause of a disease.24 This question ought to be tested by the Daubert and Hamer factors, and ought to depend on the facts of each case.
III. The Court’s New General Causation Rule
But the Court’s opinion has something to trouble everyone. While defendants may be surprised to learn that differential diagnoses are presumptively reliable, plaintiffs are now informed they must prove both general and specific causation in all toxic tort cases. Neither party asked for this; indeed, they never mention “general causation” in their four briefs. The Texas Supreme Court has never stated such a rule; Hamer notes only that plaintiffs “sometimes” offer general causation evidence when they cannot present reliable evidence of specific causation.25 In this case, Anderson made no such offer, because there were no epidemiological studies that would allow him to do so. I would not apply standards to his case that he never attempted to meet.
It is the work of a moment for a defendant to file a no-evidence motion demanding proof of general causation. The Court’s opinion short-circuits any argument that such proof is not required in every toxic tort case. General causation usually means epidemiological studies-not *617just one, but several.26 These do not exist with respect to many toxins and torts, and are very expensive to create.
Moreover, the term “toxic torts” covers a lot of ground-from asbestos and breast implants to refinery explosions, vaccinations, and food poisoning. There is a logical appeal to requiring every plaintiff to prove that an alleged causation chain can actually occur. But if asbestos fibers are directly found in a plaintiffs lungs, or thirty people eating at the same restaurant get sick at once, I hesitate to hold that each of them must prove “general causation” simply because their tort implicates a toxin.
And why limit this proclamation to toxic tort cases? “It would be a mistake to argue that the causal issues in toxic tort cases are fundamentally different from those presented in other tort cases.”27 Why not require plaintiffs to prove general causation in auto accident and slip-and-fall cases, too?
I hope the Court’s new rule is correct, but it may prove to be very expensive for some litigants if it is not. Without considering the circumstances in all of these cases, and without either party asking us to take such a step, I would not go that far here.
IV. The Dissent and the Documents
The primary issue the parties briefed and argued was whether, even without Dr. Miller’s testimony, scattered references in the medical records and a product sheet about naphtha were enough to establish causation in this case.28 The dissent would affirm on this basis — having guarded the gate and found Anderson’s expert wanting, they would permit him to proceed to verdict with documents as his only escort. The majority disagrees, but hardly explains why.
There is no question these documents were admissible; they just were not enough.29 The primary reason is that they cannot independently meet the Daubert and Havner factors that serve as a predicate for reliability. None of the notations in Anderson’s hospital records indicate the methodology used. None indicate support from medical studies or general acceptance in the medical community. Only one indicates how or why the author reached his conclusion, and that one relies only on anecdotal evidence.30 Most do not even *618mention naphtha or BOOP. If all that is needed to avoid the standards set by Dau-bert and Havner is a note in the medical records that “A caused B,” then there has been much ado about nothing.
Similarly, the naphtha product safety data sheet states only that overexposure to naphtha can cause (among many other problems) inflammation of the lungs. Because it does not mention BOOP, it provides no information about relative risk, required exposure level, or time of onset.31 This is not enough to prove causation.32
In addition, Anderson’s argument fails because it violates a rule much older than Dauberb — unless medical causation is within the common knowledge of laymen, expert testimony is required.33 There are several reasons why documents cannot replace an expert.
First, none of them are under oath. If Anderson had requested a trial continuance due to health problems, his doctor’s opinions would have to have been sworn.34 It is hard to believe his doctor’s opinion on causation — the critical issue supporting his million-dollar verdict — requires something less.
Second, it is too easy to reach the wrong conclusion by picking and choosing parts of a document and using them out of eon-text. Anderson points to several concluso-ry statements in the medical records that his lung inflammation was caused by his chemical exposure. But he skips over the discharge summary, which relates his shortness of breath to “exposure to Benzine [sic], Toluene, asbestos and naphtha.” As these entries show, hospital records often recite a working diagnosis or a patient’s version of his medical history; this does not mean they are intended to be independent opinions of causation.35 Nor does the context of such notes indicate they are based on reasonable medical probability.36
Third, Anderson’s use of these documents shows the wisdom of the rule that bars admission of “learned treatises” in place of expert testimony.37 By applying the reasoning of lawyers to the language of doctors, Anderson says we should adopt the following reasoning:
• certain chemicals (but not naphtha) have been shown to cause lung inflammation that can cause BOOP;
• naphtha can cause lung inflammation;
• therefore, naphtha-related lung inflammation must also cause BOOP.
But according to this reasoning, all chemicals that cause inflammation (and there *619are many) must cause BOOP. The medical evidence is quite the contrary.
If we drop the requirement for an escorting expert, we will soon end up with “science” that scientists can prove wrong. For example, Coastal’s expert, Dr. Keith Wilson, testified without contradiction that tobacco (for once) is not associated with this lung disease. Of course, tobacco smoke contains chemicals, which undoubtedly can inflame the lungs. Thus, tobacco must cause BOOP — except that it doesn’t.
Conclusion
Someday, medical science may find that naphtha exposure causes BOOP. But as far as we know now, it is just as likely to find that it does not. This case cannot be indefinitely postponed in the interim. In our system of justice, if a fact cannot be proved one way or the other, the party bearing the burden of proof loses. We cannot be hasty to impose liability when scientifically reliable evidence is unavailable.38
Similarly, someday our highest courts may decide all toxic tort plaintiffs must prove both general and specific causation, and that a proper differential diagnosis always proves the latter but never the former. But unlike this appeal, those questions do not have to be decided today. In our system of justice, courts should answer only the questions before them, and leave broader pronouncements to future cases. We cannot be hasty to impose rules that may require backpedaling later.
For these reasons, I concur in the judgment of the Court.
. 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993).
. 953 S.W.2d 706 (Tex.1997).
. See Daubert, 509 U.S. at 591-94, 113 S.Ct. at 2795-97; Havner, 953 S.W.2d at 711-14; Kelly v. State, 824 S.W.2d 568, 573 (Tex.Crim.App.1992).
. Literally, "after this, therefore because of this.” See Black’s Law Dictionary 1186 (7th ed.1999).
. See Black v. Food Lion, Inc., 171 F.3d 308, 313 (5th Cir.1999) (finding unreliable doctor’s conclusion that, because fall preceded fibro-myalgia, it must have caused it); Moore v. Ashland Chem., Inc., 151 F.3d 269, 278 (5th Cir.1998) (holding temporal connection between exposure and symptoms is generally insufficient absent scientific connection).
. Stedman’s Medical Dictionary 492 (27th ed.2000).
. Dr. Wilson testified without contradiction that "to really make a clear diagnosis of BOOP, you need a piece of tissue. You need a biopsy.... And so it is entirely probable that BOOP is more common than we think, but we wouldn’t make that diagnosis. If the patient gets well, which happens most of the time with BOOP, anyway, if the patient gets well, we wouldn’t do a biopsy, and so we might not know that that was specifically the diagnosis.”
. It is true doctors sometimes must find not just the disease but its cause — if a tumor is causing headaches, diagnosing the headaches does not go far enough. But Dr. Miller’s conclusion that the naphtha cargo was the cause of Anderson’s BOOP had no effect on treatment — Anderson had already left the boat.
. I make no implication about Anderson’s truthfulness or motivation. Our job is not decide his reliability, but that of the science employed by his testifying expert. Scientists take into account whether a theory was developed for litigation or something else; the Court fails to follow their rules.
. The Court says Dr. Miller’s diagnosis relied mostly on medical tests, but that is not what he said:
Q: I take it then you have to rely almost exclusively then on the history that’s given to you by the patient in order to make a judgment as to what caused it?
A: Correct
This is not to fault Dr. Miller — no laboratory test could show all the viruses, chemical fumes, or medications Anderson had been exposed to. More telling is the Court’s fallback position — diagnoses based on what patients say are reliable because there is a hearsay exception for such statements. See TexR. Evid. 803(4). Apparently one can meet the reliability requirements of Daubert and Havner simply by telling a doctor or making an excited utterance.
. See In re Paoli R.R. Yard PCB Litig., 35 F.3d 717, 762 (3d Cir.1994) (upholding exclusion of two experts who based their conclusions solely on plaintiff's self-report of illness); Viterbo v. Dow Chem. Co., 826 F.2d 420, 424 (5th Cir.1987) (holding unreliable doctor's causation opinion based solely on claimant’s oral history, stating it was no more than claimant’s testimony "dressed up and sanctified as the opinion of an expert”).
. "[Cjourts must be 'especially skeptical' of scientific evidence that has not been published or subjected to peer review.” Havner, 953 S.W.2d at 727 (citations omitted).
. See E.I. du Pont de Nemours & Co. v. Robinson, 923 S.W.2d 549, 559 (Tex.1995) (holding expert’s self-serving statement that his methodology was generally accepted in the field was insufficient to establish reliability).
. Havner, 953 S.W.2d at 717.
. See Robinson, 923 S.W.2d at 558-59 (holding expert's testimony unreliable because he did not exclude other possible causes)
. The Court's suggestion that I would reject all differential diagnoses "when applied to an illness with some unknown causes” is a straw man. I reach only the question whether a diagnosis is reliable when the most frequent cause is "unknown.”
. See Black, 171 F.3d at 313 (holding differential diagnosis insufficient to support verdict when cause of disease was unknown).
. Id. at 311 (stating flexibility was not intended "to grant open season on the admission of expert testimony by permitting courts discretionarily to disavow the Daubert factors”).
.Indeed, the Court declares “[o]ur role is not to determine reliability at all." But the Supreme Court in Havner stated, "The issue before us, as in most of the previously cited Bendectin cases, is whether the Havners' evidence is scientifically reliable and thus some evidence to support the judgment in their favor.” Havner, 953 S.W.2d at 711. Apparently, I am not alone in mistaking our role. While I agree we cannot substitute our judgment for that of the trial court, we cannot substitute our judgment for that of the Supreme Court either.
. See Black, 171 F.3d at 312 n. 2 (noting that some illnesses appear to rise and fall in prevalence depending on availability of compensation for them).
. For example, suppose a plaintiff suffering from AIDS brings suit against a health care provider alleging a transfusion of tainted blood, and denies any intravenous drug use or unprotected sexual activity. After today, this is enough to support a verdict, as AIDS can be contracted this way (general causation), and a doctor’s "differential diagnosis” based on oral history can prove it was (specific causation). Apparently, an expert need not address the very long odds against such an occurrence.
. The parties did not "rely heavily” on cases discussing differential diagnosis. In their four briefs, each party dedicated one paragraph to this issue. Neither mentioned the cases the Court adopts. Coastal’s brief merely argued that Dr. Miller never performed a differential diagnosis.
. Id. at 314.
. See Kumho Tire Co. v. Carmichael, 526 U.S. 137, 153-54, 119 S.Ct. 1167, 1177, 143 L.Ed.2d 238 (1999) (holding issue was not whether it was reasonable in general to use sight and touch to test tire's tread, but whether it was reasonable to use them to draw the conclusions involved in that case).
. Havner, 953 S.W.2d at 715.
. See id. at 727. Havner rejected in vivo and in vitro studies offered in that case on grounds that make them unlikely to be more persuasive in other cases. Id. at 728-30.
. Joseph Sanders & Julie Machal Fulks, The Admissibility of Differential Diagnosis Testimony to Prove Causation in Toxic Tort Cases: The Interplay of Adjective and Substantive Law, 64 Law & Contemp. Probs. 107, 110 (2001).
. Anderson also argues circumstantial evidence supports causation, relying on the temporal connection between his exposure and his BOOP. For the reasons noted above, this connection is not enough alone to meet Dau-bert — with or without an expert.
. The dissent suggests Coastal waived error by not itemizing its objections to each line in the medical records on Robinson/Havner grounds. But Coastal specifically objected to all causation opinions in the medical records as unreliable, and challenged all of Anderson’s proof of causation by filing a pretrial motion, urging that motion again before trial began, objecting repeatedly at trial, moving twice for a directed verdict, objecting to the relevant portions of the jury charge, moving for judgment notwithstanding the verdict, and moving for a new trial. The trial court could not have missed Coastal's complaint. See TexJR.App. P. 33.1(a)(1).
.Dr. Brown's admission note states he based his conclusion on his experience as a medical director at a chemical plant, which made him familiar with “the types of pneumonia that this particular product [naphtha] could cause.” This is exactly the kind of *618anecdotal evidence that, while it may appear convincing to a particular practitioner, proves nothing because it has no scientific basis. See Havner, 953 S.W.2d at 719-20.
. See Havner, 953 S.W.2d at 720.
. See Moore, 151 F.3d at 278 (holding material safety data sheet could not support expert’s causation opinion when expert did not know what tests it was based on or whether plaintiff's exposure level was included).
. Leitch v. Hornsby, 935 S.W.2d 114, 119 (Tex.1996) (holding plaintiff must provide probative evidence through expert testimony connecting injury to tortious act); see also Bowles v. Bourdon, 148 Tex. 1, 219 S.W.2d 779, 782 (1949).
. Tex.R. Civ. P. 251; Hawthorne v. Guenther, 917 S.W.2d 924, 930 (Tex.App.-Beaumont 1996, writ denied).
. Burroughs Wellcome Co. v. Crye, 907 S.W.2d 497, 500 (Tex.1995).
. Id.
. Tex.R. Evid. 803(18) (providing statement from learned document may be read into evidence-though not received as an exhibit — if an expert is present to verify its reliability and explain what it means).
. Havner, 953 S.W.2d at 728.