[920] Justice Breyer
delivered the opinion of the Court.
We again consider the right to an abortion. We understand the controversial nature of the problem. Millions of Americans believe that life begins at conception and consequently that an abortion is akin to causing the death of an innocent child; they recoil at the thought of a law that would permit it. Other millions fear that a law that forbids abortion would condemn many American women to lives that lack dignity, depriving them of equal liberty and leading those with least resources to undergo illegal abortions with the attendant risks of death and suffering. Taking account of [921] these virtually irreconcilable points of view, aware that constitutional law must govern a society whose different members sincerely hold directly opposing views, and considering the matter in light of the Constitution’s guarantees of fundamental individual liberty, this Court, in the course of a generation, has determined and then redetermined that the Constitution offers basic protection to the woman’s right to choose. Roe v. Wade, 410 U. S. 113 (1973); Planned Parenthood of Southeastern Pa. v. Casey, 505 U. S. 833 (1992). We shall not revisit those legal principles. Rather, we apply them to the circumstances of this case.
Three established principles determine the issue before us. We shall set them forth in the language of the joint opinion in Casey. First, before "viability... the woman has a right to choose to terminate her pregnancy.” Id., at 870 (plurality opinion).
Second, "a law designed to further the State’s interest in fetal life which imposes an undue burden on the woman’s decision before fetal viability” is unconstitutional. Id., at 877. An "undue burden is ... shorthand for the conclusion that a state regulation has the purpose or effect of placing a substantial obstacle in the path of a woman seeking an abortion of a nonviable fetus.” Ibid.
Third, “ ‘subsequent to viability, the State in promoting its interest in the potentiality of human life may, if it chooses, regulate, and even proscribe, abortion except where it is necessary, in appropriate medical judgment, for the preservation of the life or health of the mother.’ ” Id., at 879 (quoting Roe v. Wade, supra, at 164-165).
We apply these principles to a Nebraska law banning “partial birth abortion.” The statute reads as follows:
“No partial birth abortion shall be performed in this state, unless such procedure is neeessary to save the life of the mother whose life is endangered by a physical disorder, physical illness, or physical injury, including a life-endangering physical condition caused by or aris[922] ing from the pregnancy itself.” Neb. Rev. Stat. Ann. §28-328(1) (Supp. 1999).
The statute defines “partial birth abortion” as:
“an abortion procedure in which the person performing the abortion partially delivers vaginally a living unborn child before killing the unborn child and completing the delivery.” § 28-326(9).
It further defines “partially delivers vaginally a living unborn child before killing the unborn child” to mean
“deliberately and intentionally delivering into the vagina a living unborn child, or a substantial portion thereof, for the purpose of performing a procedure that the person performing such procedure knows will kill the unborn child and does kill the unborn child.” Ibid.
The law classifies violation of the statute as a “Class III felony” carrying a prison term of up to 20 years, and a fine of up to $25,000. §§28-328(2), 28-105. It also provides for the automatic revocation of a doctor’s license to practice medicine in Nebraska. § 28-328(4).
We hold that this statute violates the Constitution.
HH
<j
Dr. Leroy Carhart is a Nebraska physician who performs abortions in a clinical setting. He brought this lawsuit in Federal District Court seeking a declaration that the Nebraska statute violates the Federal Constitution, and asking for an injunction forbidding its enforcement. After a trial on the merits, during which both sides presented several expert witnesses, the District Court held the statute unconstitutional. 11F. Supp. 2d 1099 (Neb. 1998). On appeal, the Eighth Circuit affirmed. 192 F. 3d 1142 (1999); cf. Hope Clinic v. Ryan, 195 F. 3d 857 (CA7 1999) (en banc) (consider[923] ing a similar statute, but reaching a different legal conclusion). We granted certiorari to consider the matter.
B
Because Nebraska law seeks to ban one method of aborting a pregnancy, we must describe and then discuss several different abortion procedures. Considering the fact that those procedures seek to terminate a potential human life, our discussion may seem clinically cold or callous to some, perhaps horrifying to others. There is no alternative way, however, to acquaint the reader with the technical distinctions among different abortion methods and related factual matters, upon which the outcome of this ease depends. For that reason, drawing upon the findings of the trial court, underlying testimony, and related medical texts, we shall describe the relevant methods of performing abortions in technical detail.
The evidence before the trial court, as supported or supplemented in the literature, indicates the following:
1. About 90% of all abortions performed in the United States take place during the first trimester of pregnancy, before 12 weeks of gestational age. Centers for Disease Control and Prevention, Abortion Surveillance — United States, 1996, p. 41 (July 30, 1999) (hereinafter Abortion Surveillance). During the first trimester, the predominant abortion method is “vacuum aspiration," which involves insertion of a vacuum tube (cannula) into the uterus to evacuate the contents. Such an abortion is typically performed on an outpatient basis under local anesthesia. 11 F. Supp. 2d, at 1102; Obstetrics: Normal & Problem Pregnancies 1253-1254 (S. Gabbe, J. Niebyl, & J. Simpson eds. 3d ed. 1996). Vacuum aspiration is considered particularly safe. The procedure’s mortality rates for first trimester abortion ai'e, for example, 5 to 10 times lower than those associated with carrying the fetus to term. Complication rates are also low. Id., at 1251; Lawson et al., Abortion Mortality, United [924] States, 1972 through 1987, 171 Am. J. Obstet. Gynecol. 1365, 1368 (1994); M. Paul et al., A Clinicians Guide to Medical and Surgical Abortion 108-109 (1999) (hereinafter Medical and Surgical Abortion). As the fetus grows in size, however, the vacuum aspiration method becomes increasingly difficult to use. 11 F. Supp. 2d, at 1102-1103; Obstetrics: Normal & Problem Pregnancies, supra, at 1268.
2. Approximately 10% of all abortions are performed during the second trimester of pregnancy (12 to 24 weeks). Abortion Surveillance 41. In the early 1970’s, inducing labor through the injection of saline into the uterus was the predominant method of second trimester abortion. Id., at 8; Planned Parenthood of Central Mo. v. Danforth, 428 U. S. 52, 76 (1976). Today, however, the medical profession has switched from medical induction of labor to surgical procedures for most second trimester abortions. The most commonly used procedure is called “dilation and evacuation” (D&E). That procedure (together with a modified form of vacuum aspiration used in the early second trimester) accounts for about 95% of all abortions performed from 12 to 20 weeks of gestational age. Abortion Surveillance 41.
3. D&E “refers generically to transcervieal procedures performed at 13 weeks gestation or later.” American Medical Association, Report of Board of Trustees on Late-Term Abortion, App. 490 (hereinafter AMA Report). The AMA Report, adopted by the District Court, describes the process as follows.
Between 13 and 15 weeks of gestation:
“D&E is similar to vacuum aspiration except that the cervix must be dilated more widely because surgical instruments are used to remove larger pieces of tissue. Osmotic dilators are usually used. Intravenous fluids and an analgesic or sedative may be administered. A local anesthetic such as a paracervical block may be administered, dilating agents, if used, are removed and instruments are inserted through the cervix into the [925] uterus to removal fetal and placental tissue. Because fetal tissue is friable and easily broken, the fetus may not be removed intact. The walls of the uterus are scraped with a curette to ensure that no tissue remains.” Id., at 490-491.
After 15 weeks:
“Because the fetus is larger at this stage of gestation (particularly the head), and because bones are more rigid, dismemberment or other destructive procedures are more likely to be required than at earlier gestational ages to remove fetal and placental tissue.” Id., at 491.
After 20 weeks:
“Some physicians use intrafetal potassium chloride or digoxin to induce fetal demise prior to a late D&E (after 20 weeks), to facilitate evacuation.” Id., at 491-492.
There are variations in D&E operative strategy; compare ibid, with W. Hern, Abortion Practice 146-156 (1984), and Medical and Surgical Abortion 133-185. However, the common points are that D&E involves (1) dilation of the cervix; (2) removal of at least some fetal tissue using nonvacuum instruments; and (3) (after the 15th week) the potential need for instrumental disarticulation or dismemberment of the fetus or the collapse of fetal parts to facilitate evacuation from the uterus.
4. When instrumental disarticulation incident to D&E is necessary, it typically occurs as the doctor pulls a portion of the fetus through the cervix into the birth canal. Dr. Car-hart testified at trial as follows:
“Dr. Carhart:.. . ‘The dismemberment occurs between the traction of ... my instrument and the counter-traction of the internal os of the cervix....
“Counsel: ‘So the dismemberment occurs after you pulled a part of the fetus through the cervix, is that correct?
[926] “Dr. Carhart: ‘Exactly. Because you’re using — The cervix has two strictures or two rings, the internal os and the external os . . . that’s what’s actually doing the dismembering....
“Counsel: ‘When we talked before or talked before about a D&E, that is not — where there is not intention to do it intact, do you, in that situation, dismember the fetus in útero first, then remove portions?
“Dr. Carhart: T don’t think so.... I don’t know of any way that one could go in and intentionally dismember the fetus in the uterus. ... It takes something that restricts the motion of the fetus against what you’re doing before you’re going to get dismemberment.’ ” 11 F. Supp. 2d, at 1104.
Dr. Carhart’s specification of the location of fetal disarticulation is consistent with other sources. See Medical and Surgical Abortion 135; App. in Nos. 98-3245 and 98-3300 (CA8), p. 683, (testimony of Dr. Phillip Stubblefield) (“Q: So you don’t actually dismember the fetus in útero, then take the pieces out? A: No”).
5. The D&E procedure carries certain risks. The use of instruments within the uterus creates a danger of accidental perforation and damage to neighboring organs. Sharp fetal bone fragments create similar dangers. And fetal tissue accidentally left behind can cause infection and various other complications. See 11F. Supp. 2d, at 1110; Gynecologic, Obstetric, and Related Surgery 1045 (D. Nichols & D. Clarke-Pearson eds. 2d ed. 2000); F. Cunningham et al., Williams Obstetrics 598 (20th ed. 1997). Nonetheless studies show that the risks of mortality and complication that accompany the D&E procedure between the 12th and 20th weeks of gestation are significantly lower than those accompanying induced labor procedures (the next safest midseeond trimester procedures). See Gynecologic, Obstetric, and Related Surgery, supra, at 1046; AMA Report, App. 495, 496; Medical [927] and Surgical Abortion 139,142; Lawson, 171 Am. J. Obstet. Gynecol., at 1368.
6. At trial, Dr. Carhart and Dr. Stubblefield described a variation of the D&E procedure, which they referred to as an “intact D&E.” See 11 F. Supp. 2d, at 1105, lili. Like other versions of the D&E technique, it begins with induced dilation of the cervix. The procedure then involves removing the fetus from the uterus through the cervix “intact,” i. e., in one pass, rather than in several passes. Ibid. It is used after 16 weeks at the earliest, as vacuum aspiration becomes ineffective and the fetal skull becomes too large to pass through the cervix. Id., at 1105. The intact D&E proceeds in one of two ways, depending on the presentation of the fetus. If the fetus presents head first (a vertex presentation), the doctor collapses the skull; and the doctor then extracts the entire fetus through the cervix. If the fetus presents feet first (a breech presentation), the doctor pulls the fetal body through the cervix, collapses the skull, and extracts the fetus through the cervix. Ibid. The breech extraction version of the intact D&E is also known commonly as “dilation and extraction,” or D&X. Id., at 1112. In the late second trimester, vertex, breech, and traverse/ compound (sideways) presentations occur in roughly similar proportions. Medical and Surgical Abortion 135; 11F. Supp. 2d, at 1108.
7. The intact D&E procedure can also be found described in certain obstetric and abortion clinical textbooks, where two variations are recognized. The first, as just described, calls for the physician to adapt his method for extracting the intact fetus depending on fetal presentation. See Gynecologic, Obstetric, and Related Surgery, supra, at 1043; Medical and Surgical Abortion 136-137. This is the method used by Dr. Carhart. See 11F. Supp. 2d, at 1105. A slightly different version of the intact D&E procedure, associated with Dr. Martin Haskell, calls for conversion to a breech presentation in all cases. See Gynecologic, Obstetric, and Related [928] Surgery, supra, at 1043 (citing M. Haskell, Dilation and Extraction for Late Second Trimester Abortion (1992), in 139 Cong. Ree. 8605 (1993)).
8. The American College of Obstetricians and Gynecologists describes the D&X procedure in a manner corresponding to a breech-conversion intact D&E, including the following steps:
"1. deliberate dilatation of the cervix, usually over a sequence of days;
“2. instrumental conversion of the fetus to a footling breech;
“3. breech extraction of the body excepting the head; and
“4. partial evacuation of the intracranial contents of a living fetus to effect vaginal delivery of a dead but otherwise intact fetus.” American College of Obstetricians and Gynecologists Executive Board, Statement on Intact Dilation and Extraction (Jan. 12, 1997) (hereinafter ACOG Statement), App. 599-560.
Despite the technical differences we have just described, intact D&E and D&X are sufficiently similar for us to use the terms interchangeably.
9. Dr. Carhart testified he attempts to use the intact D&E procedure during weeks 16 to 20 because (1) it reduces the dangers from sharp bone fragments passing through the cervix, (2) minimizes the number of instrument passes needed for extraction and lessens the likelihood of uterine perforations caused by those instruments, (3) reduces the likelihood of leaving infection-causing fetal and placental tissue in the uterus, and (4) could help to prevent potentially fatal absorption of fetal tissue into the maternal circulation. See 11 F. Supp. 2d, at 1107. The District Court made no findings about the D&X procedure’s overall safety. Id., at 1126, n. 39. The District Court concluded, however, that “the evidence is both clear and convincing that Carhart’s [929] D&X procedure is superior to, and safer than, the ... other abortion procedures used during the relevant gestational period in the 10 to 20 eases a year that present to Dr. Garhart ” Id., at 1126.
10. The materials presented at trial referred to the potential benefits of the D&X procedure in circumstances involving nonviable fetuses, such as fetuses with abnormal fluid accumulation in the brain (hydrocephaly). See 11 F. Supp. 2d, at 1107 (quoting AMA Report, App. 492 (“ ‘Intact D&X may be preferred by some physicians, particularly when the fetus has been diagnosed with hydrocephaly or other anomalies incompatible with life outside the womb’ ”)); see also Grimes, The Continuing Need for Late Abortions, 280 JAMA 747, 748 (Aug. 26, 1998) (D&X “may be especially useful in the presence of fetal anomalies, such as hydrocephalus,” because its reduction of the cranium allows “a smaller diameter to pass through the cervix, thus reducing risk of cervical injury”). Others have emphasized its potential for women with prior uterine sears, or for women for whom induction of labor would be particularly dangerous. See Women’s Medical Professional Corp. v. Voinovich, 911 F. Supp. 2d 1051, 1067 (SD Ohio 1995); Evans v. Kelley, 977 F. Supp. 2d 1283, 1296 (ED Mich. 1997).
11. There are no reliable data on the number of D&X abortions performed annually. Estimates have ranged between 640 and 5,000 per year. Compare Henshaw, Abortion Incidence and Services in the United States, 1995-1996,30 Family Planning Perspectives 263,268 (1998), with Joint Hearing on S. 6 and H. R. 929 before the Senate Committee on the Judiciary and the Subcommittee on the Constitution of the House Committee on the Judiciary, 105th Cong., 1st Sess., 46 (1997).
II
The question before us is whether Nebraska’s statute, making criminal the performance of a “partial birth abortion,” violates the Federal Constitution, as interpreted in [930] Planned Parenthood of Southeastern Pa. v. Casey, 505 U. S. 833 (1992), and Roe v. Wade, 410 U. S. 113 (1973). We conclude that it does for at least two independent reasons. First, the law lacks any exception “ ‘for the preservation of the... health of the mother.’ ” Casey, 505 U. S., at 879 (plurality opinion). Second, it “imposes an undue burden on a woman’s ability” to choose a D&E abortion, thereby unduly burdening the right to choose abortion itself. Id., at 874. We shall discuss each of these reasons in turn.
A
The Casey plurality opinion reiterated what the Court held in Roe; that “ ‘subsequent to viability, the State in promoting its interest in the potentiality of human life may, if it chooses, regulate, and even proscribe, abortion except where it is necessary, in appropriate medical judgment, for the preservation of the life or health of the mother.’” 505 U. S., at 879 (quoting Roe, supra, at 164-165) (emphasis added).
The fact that Nebraska’s law applies both previability and postviability aggravates the constitutional problem presented. The State’s interest in regulating abortion previa-bility is considerably weaker than postviability. See Casey, supra, at 870. Since the law requires a health exception in order to validate even a postviability abortion regulation, it at a minimum requires the same in respect to previability regulation. See Casey, supra, at 880 (majority opinion) (assuming need for health exception previability); see also Harris v. McRae, 448 U. S. 297, 316 (1980).
The quoted standard also depends on the state regulations “promoting [the State’s] interest in the potentiality of human life.” The Nebraska law, of course, does not directly further an interest “in the potentiality of human life” by saving the fetus in question from destruction, as it regulates only a method of performing abortion. Nebraska describes its interests differently It says the law “ ‘show[s] concern for the life of the unborn,’ ” “prevents] cruelty to partially born chil[931] dren,” and “preserved] the integrity of the medical profession.” Brief for Petitioners 48. But we cannot see how the interest-related differences could make any difference to the question at hand, namely, the application of the “health” requirement.
Consequently, the governing standard requires an exception “where it is necessary, in appropriate medical judgment for the preservation of the life or health of the mother,” Casey, supra, at 879, for this Court has made clear that a State may promote but not endanger a woman’s health when it regulates the methods of abortion. Thornburgh v. American College of Obstetricians and Gynecologists, 476 U. S. 747, 768-769 (1986); Colautti v. Franklin, 439 U. S. 379, 400 (1979); Danforth, 428 U. S., at 76-79; Doe v. Bolton, 410 U. S. 179, 197 (1973).
Justice Thomas says that the cases just cited limit this principle to situations where the pregnancy itself creates a threat to health. See post, at 1010. He is wrong. The cited eases, reaffirmed in Casey, recognize that a State cannot subject women’s health to significant risks both in that context, and also where state regulations force women to use riskier methods of abortion. Our cases have repeatedly invalidated statutes that in the process of regulating the methods of abortion, imposed significant health risks. They make clear that a risk to a women’s health is the same whether it happens to arise from regulating a particular method of abortion, or from barring abortion entirely. Our holding does not go beyond those cases, as ratified in Casey.
Nebraska responds that the law does not require a health exception unless there is a need for such an exception. And here there is no such need, it says. It argues that “safe alternatives remain available” and “a ban on partial-birth abortion/D&X would create no risk to the health of women.” Brief for Petitioners 29, 40. The problem for Nebraska is [932] that the parties strongly contested this factual question in the trial court below; and the findings and evidence support Dr. Carhart. The State fails to demonstrate that banning D&X without a health exception may not create significant health risks for women, because the record shows that significant medical authority supports the proposition that in some circumstances, D&X would be the safest procedure.
We shall reiterate in summary form the relevant findings and evidence. On the basis of medical testimony the District Court concluded that “Carhart’s D&X procedure is ... safer tha[n] the D&E and other abortion procedures used during the relevant gestational period in the 10 to 20 eases a year that present to Dr. Carhart.” 11F. Supp. 2d, at 1126. It found that the D&X procedure permits the fetus to pass through the cervix with a minimum of instrumentation. Ibid. It thereby
“reduces operating time, blood loss and risk of infection; reduces complications from bony fragments; reduces instrument-inflicted damage to the uterus and cervix; prevents the most common causes of maternal mortality (DIC and amniotie fluid embolus); and eliminates the possibility of ‘horrible complications’ arising from retained fetal parts.” Ibid.
The District Court also noted that a select panel of the American College of Obstetricians and Gynecologists concluded that D&X “ ‘may be the best or most appropriate procedure in a particular circumstance to save the life or preserve the health of a woman.’ ” Id., at 1105, n. 10 (quoting ACOG Statement, App. 600-601) (but see an important qualification, infra, at 934). With one exception, the federal trial courts that have heard expert evidence on the matter have reached similar factual conclusions. See Rhode Island Medical Soc. v. Whitehouse, 66 F. Supp. 2d 288, 314 (RI . 1999); A Choice for Women v. Butterworth, 54 F. Supp. 2d 1148, 1153, 1156 (SD Fla. 1998); Causeway Medical Suite v. Foster, 48 F. Supp. 2d 604, 613-614 (ED La. 1999); Richmond [933] Medical Center for Women v. Gilmore, 11 F. Supp. 2d 795, 827, n. 40 (ED Va. 1998); Hope Clinic v. Ryan, 995 F. Supp. 2d 847, 852 (ND Ill. 1998), vacated, 195 F. 8d 857 (CA7 1999), cert. pending, No. 99-1152; Voinovich, 911 F. Supp. 2d, at 1069-1070; Kelley, 977 F. Supp. 2d, at 1296; but see Planned Parenthood of Wis. v. Doyle, 44 F. Supp. 2d 975, 980 (WD Wis.), vacated, 195 F. 3d 857 (CA7 1999).
Nebraska, along with supporting amici, replies that these findings are irrelevant, wrong, or applicable only in a tiny number of instances. It says (1) that the D&X procedure is “little-used,” (2) by only “a handful of doctors.” Brief for Petitioners 32. It argues (3) that D&E and labor induction are at all times “safe alternative procedures.” Id., at 36. It refers to the testimony of petitioners’ medical expert, who testified (4) that the ban would not increase a woman’s risk of several rare abortion complications (disseminated intra-vascular coagulopathy and amniotie fluid embolus), id., at 37; App. 642-644.
The Association of American Physicians and Surgeons et al., amici supporting Nebraska, argue (5) that elements of the D&X procedure may create special risks, including cervical incompetence caused by overdilitation, injury caused by conversion of the fetal presentation, and dangers arising from the “blind” use of instrumentation to pierce the fetal skull while lodged in the birth canal. See Brief for Association of American Physicians and Surgeons et al. as Amici Curiae 21-23; see also Sprang & Neerhof, Rationale for Banning Abortions Late in Pregnancy, 280 JAMA 744,746 (Aug. 26,1998).
Nebraska further emphasizes (6) that there are no medical studies “establishing the safety of the partial-birth abortion/ D&X procedure,” Brief for Petitioners 39, and “no medical studies comparing the safety of partial-birth abortion/D&X to other abortion procedures,” ibid. It points to, id., at 35, [934] (7) an American Medical Association policy statement that “‘there does not appear to be any identified situation in which intact D&X is the only appropriate procedure to induce abortion,’” Late Term Pregnancy Termination Techniques, AMA Policy H-5.982 (1997). And it points out (8) that the American College of Obstetricians and Gynecologists qualified its statement that D&X “may be the best or most appropriate procedure,” by adding that the panel “could identify no circumstances under which [the D&X] procedure . . . would be the only option to save the life or preserve the health of the woman.” App. 600-601.
We find these eight arguments insufficient to demonstrate that Nebraska’s law needs no health exception. For one thing, certain of the arguments are beside the point. The D&X procedure’s relative rarity (argument (1)) is not highly relevant. The D&X is an infrequently used abortion procedure; but the health exception question is whether protecting women’s health requires an exception for those infrequent occasions. A rarely used treatment might be necessary to treat a rarely occurring disease that could strike anyone — the State cannot prohibit a person from obtaining treatment simply by pointing out that most people do not need it. Nor can we know whether the fact that only a “handful” of doctors use the procedure (argument (2)) reflects the comparative rarity of late second term abortions, the procedure’s recent development, Gynecologic, Obstetric, and Related Surgery, at 1043, the controversy surrounding it, or, as Nebraska suggests, the procedure’s lack of utility.
For another thing, the record responds to Nebraska’s (and amici’s) medically based arguments. In respect to argument (3), for example, the District Court agreed that alternatives, such as D&E and induced labor, are “safe” but found that the D&X method was significantly safer in certain circumstances. 11 F. Supp. 2d, at 1125-1126. In respect to [935] argument (4), the District Court simply- relied on different expert testimony — testimony stating that “ ‘[a]nother advantage of the Intact D&E is that it eliminates the risk of embolism of cerebral tissue into the woman’s blood stream.’ ” Id., at 1124 (quoting Hearing on H. R. 1833 before the Senate Committee on the Judiciary, 104th Cong., 1st Sess., 260 (1995) (statement of W. Hern).
In response to amici’s argument (5), the American College of Obstetricians and Gynecologists, in its own amici brief, denies that D&X generally poses risks greater than the alternatives. It says that the suggested alternative procedures involve similar or greater risks of cervical and. uterine injury, for “D&E procedures, involve similar amounts of dili-tation” and “of course childbirth involves even greater cervical dilitation.” Brief for American College of Obstetricians and Gynecologists et al. as Amici Curiae 23. The College points out that Dr. Carhart does not reposition the fetus thereby avoiding any risks stemming from conversion to breech presentation, and that, as compared with D&X, D&E involves the same, if not greater, “blind” use of sharp instruments in the uterine cavity. Id., at 23-24.
We do not quarrel with Nebraska’s argument (6), for Nebraska is right. There are no general medical studies documenting comparative safety. Neither do we deny the import of the American Medical Association’s statement (argument (7)) — even though the State does omit the remainder of that statement: “The AMA recommends that the procedure not be used unless alternative procedures pose materially greater risk to the woman.” Late Term Pregnancy Termination Techniques, AMA Policy H-5.982 (emphasis added).
We cannot, however, read the American College of Obstetricians and Gynecologists panel’s qualification (that it could not “identify” a circumstance where D&X was the “only” life- or health-preserving option) as if, according to Nebraska’s argument (8), it denied the potential health-related need [936] for D&X. That is because the College writes the following in its amici brief:
“Depending on the physician’s skill and experience, the D&X procedure ean be the most appropriate abortion procedure for some women in some circumstances. D&X presents a variety of potential safety advantages over other abortion procedures used during the same gestational period. Compared to D&Bs involving dismemberment, D&X involves less risk of uterine perforation or cervical laceration because it requires the physician to make fewer passes into the uterus with sharp instruments and reduces the presence of sharp fetal bone fragments that ean injure the uterus and cervix. There is also considerable evidence that D&X reduces the risk of retained fetal tissue, a serious abortion complication that ean cause maternal death, and that D&X reduces the incidence of a 'free floating’ fetal head that can be difficult for a physician to grasp and remove and ean thus cause maternal injury. That D&X procedures usually take less time than other abortion methods used at a comparable stage of pregnancy can also have health advantages. The shorter the procedure, the less blood loss, trauma, and exposure to anesthesia. The intuitive safety advantages of intact D&E are supported by clinical experience. Especially for women with particular health conditions, there is medical evidence that D&X may be safer than available alternatives.” Brief for American College of Obstetricians and Gynecologists et al. as Amici Curiae 21-22 (citation and footnotes omitted).
The upshot is a District Court finding that D&X significantly obviates health risks in certain circumstances, a highly plausible reeord-based explanation of why that might be so, a division of opinion among some medical experts over [937] whether D&X is generally safer, and an absence of controlled medical studies that would help answer these medical questions. Given these medically related evidentiary circumstances, we believe the law requires a health exception.
The word “necessary5’ in Casey’s phrase “necessary, in appropriate medical judgment, for the preservation of the life or health of the mother,” 505 U. S., at 879 (internal quotation marks omitted), cannot refer to an absolute necessity or to absolute proof. Medical treatments and procedures are often considered appropriate (or inappropriate) in light of estimated comparative health risks (and health benefits) in particular cases. Neither can that phrase require unanimity of medical opinion. Doctors often differ in their estimation of comparative health risks and appropriate treatment. And Casey’s words “appropriate medical judgment” must embody the judicial need to tolerate responsible differences of medical opinion — differences of a sort that the American Medical Association and American College of Obstetricians and Gynecologists’ statements together indicate are present here.
For another thing, the division of medical opinion about the matter at most means uncertainty, a factor that signals the presence of risk, not its absence. That division here involves highly qualified knowledgeable experts on both sides of the issue. Where a significant body of medical opinion believes a procedure may bring with it greater safety for some patients and explains the medical reasons supporting that view, we cannot say that the presence of a different view by itself proves the contrary. Rather, the uncertainty means a significant likelihood that those who believe that D&X is a safer abortion method in certain circumstances may turn out to be right. If so, then the absence of a health exception will place women at an unnecessary risk of tragic health consequences. If they are wrong, the exception will simply turn out to have been unnecessary.
In sum, Nebraska has not convinced us that a health exception is “never necessary to preserve the health of [938] women.” Reply Brief for Petitioners 4. Rather, a statute that altogether forbids D&X creates a significant health risk. The statute consequently must contain a health exception. This is not to say, as Justice Thomas and Justice Kennedy claim, that a State is prohibited from proscribing an abortion procedure whenever a particular physician deems the procedure preferable. By no means must a State grant physicians “unfettered discretion” in their selection of abortion methods. Post, at 969 (Kennedy, J., dissenting). But where substantial medical authority supports the proposition that banning a particular abortion procedure could endanger women’s health, Casey requires the statute to include a health exception when the procedure is “'necessary, in appropriate medical judgment, for the preservation of the life or health of the mother.’” 505 U.S., at 879. Requiring such an exception in this ease is no departure from Casey, but simply a straightforward application of its holding.
B
The Eighth Circuit found the Nebraska statute unconstitutional because, in Casey’s words, it has the “effect of placing a substantial obstacle in the path of a woman seeking an abortion of a nonviable fetus.” Id., at 877. It thereby places an “undue burden” upon a woman’s right to terminate her pregnancy before viability. Ibid. Nebraska does not deny that the statute imposes an “undue burden” if it applies to the more commonly used D&E procedure as well as to D&X. And we agree with the Eighth Circuit that it does so apply.
Our earlier discussion of the D&E procedure, supra, at 924-926, shows that it falls within the statutory prohibition. The statute forbids “deliberately and intentionally delivering into the vagina a living unborn child, or a substantial portion thereof, for the purpose of performing a procedure that the person performing such procedure knows will kill the unborn child.” Neb. Rev. Stat. Ann. §28-326(9) (Supp. 1999). We [939] do not understand how one could distinguish, using this language, between D&E (where a foot or arm is drawn through the cervix) and D&X (where the body up to the head is drawn through the cervix). Evidence before the trial court makes clear that D&E will often involve a physician pulling a “substantial portion” of a still living fetus, say, an arm or leg, into the vagina prior to the death of the fetus. 11 F. Supp. 2d, at 1128; id., at 1128-1130. Indeed D&E involves dismemberment that commonly occurs only when the fetus meets resistance that restricts the motion of the fetus: “The dismemberment occurs between the traction of... [the] instrument and the counter-traction of the internal os of the cervix.” Id., at 1128. And these events often do not occur until after a portion of a living fetus has been pulled into the vagina. Id., at 1104; see also Medical and Surgical Abortion 135 (“During the mid-second trimester, separation of the fetal corpus may occur when the fetus is drawn into the lower uterine segment, where compression and traction against the endoeervix facilitates disarticulation”).
Even if the statute’s basic aim is to ban D&X, its language makes clear that it also covers a much broader category of procedures. The language does not track the medical differences between D&E and D&X — though it would have been a simple matter, for example, to provide an exception for the performance of D&E and other abortion procedures. E. g., Kan. Stat. Ann. §65-6721(b)(l) (Supp. 1999). Nor does the statute anywhere suggest that its application turns on whether a'portion of the fetus’ body is drawn into the vagina as part of a process to extract an intact fetus after collapsing the head as opposed to a process that would dismember the fetus. Thus, the dissenters’ argument that the law was generally intended to bar D&X can be both correct and irrelevant. The relevant question is not whether the legislature wanted to ban D&X; it is whether the law was intended to apply only to D&X. The plain language covers both procedures. A rereading of this opinion, supra, at 924-929, as [940] well as Justice Thomas’ dissent, post, at 984-987, will make clear why we can find no difference, in terms of this statute, between the D&X procedure as described and the D&E procedure as it might be performed. (In particular, compare post, at 984-986 (THOMAS, J., dissenting), with post, at 986-989 (Thomas, J., dissenting).) Both procedures can involve the introduction of a “substantial portion” of a still living fetus, through the cervix, into the vagina — the very feature of an abortion that leads Justice Thomas to characterize such a procedure as involving “partial birth.”
The Nebraska State Attorney General argues that the statute does differentiate between the two procedures. He says that the statutory words “substantial portion” mean “the child up to the head.” He consequently denies the statute’s application where the physician introduces into the birth canal a fetal arm or leg or anything less than the entire fetal body. Brief for Petitioners 20. He argues further that we must defer to his views about the meaning of the state statute. Id., at 12-13.
We cannot accept the Attorney General’s narrowing interpretation of the Nebraska statute. This Court’s case law makes clear that we are not to give the Attorney General’s interpretative views controlling weight. For one thing, this Court normally follows lower federal-court interpretations of state law. McMillian v. Monroe County, 520 U. S. 781, 786 (1997); Brockett v. Spokane Arcades, Inc., 472 U. S. 491, 500, n. 9 (1985). It “rarely reviews a construction of state law agreed upon by the two lower federal courts.” Virginia v. American Booksellers Assn., Inc., 484 U. S. 383, 395 (1988). In this ease, the two lower courts have both rejected the Attorney General’s narrowing interpretation.
For another, our precedent warns against accepting as “authoritative” an Attorney General’s interpretation of state law when “the Attorney General does not bind the state courts or local law enforcement authorities.” Ibid. [941] Under Nebraska law, the Attorney General’s interpretative views do not bind the state courts. State v. Coffman, 213 Neb. 560, 561, 330 N. W. 2d 727, 728 (1983) (Attorney General’s issued opinions, while entitled to “substantial weight” and “to be respectfully considered,” are of “no controlling authority”). Nor apparently do they bind elected county attorneys, to whom Nebraska gives an independent authority to initiate criminal prosecutions. Neb. Rev. Stat. Ann. §§23-1201(1), 28-328(5), 84-205(3) (Supp. 1999); cf. Crandon v. United States, 494 U. S. 152, 177 (1990) (Scalia, J., concurring in judgment) (“[W]e have never thought that the interpretation of those charged with prosecuting criminal statutes is entitled to deference”).
Nor can we say that the lower courts used the wrong legal standard in assessing the Attorney General’s interpretation. The Eighth Circuit recognized its “duty to give [the law] a construction . . . that would avoid constitutional doubts.” 192 F. 3d, at 1150. It nonetheless concluded that the Attorney General’s interpretation would “twist the words of the law and give them a meaning they cannot reasonably bear.” Ibid. The Eighth Circuit is far from alone in rejecting such a narrowing interpretation. The language in question is based on model statutory language (though some States omit any further definition of “partial birth abortion”), which 10 lower federal courts have considered on the merits. All 10 of those courts (including the Eighth Circuit) have found the language potentially applicable to other abortion procedures. See Planned Parenthood of Greater Iowa, Inc. v. Miller, 195 F. 3d 386 (CA8 1999); Little Rock Family Planning Services v. Jegley, 192 F. 3d 794, 797-798 (CA8 1999); Hope Clinic, 195 F. 3d, at 865-871 (imposing precautionary injunction to prevent application beyond D&X); id., at 885-889 (Posner, C. J., dissenting); Rhode Island Medical Soc., 66 F. Supp. 2d, at 309-310; Richmond Medical Center for Women, 55 [942] F. Supp. 2d, at 471; A Choice for Women, 54 F. Supp. 2d, at 1155; Causeway Medical Suite, 43 F. Supp. 2d, at 614-615; Planned Parenthood of Central N. J. v. Verniero, 41 F. Supp. 2d 478, 503-504 (NJ 1998); Eubanks v. Stengel, 28 F. Supp. 2d 1024, 1034-1035 (WD Ky. 1998); Planned Parenthood of Southern Ariz., Inc. v. Woods, 982 F. Supp. 2d 1369, 1378 (Ariz. 1997); Kelley, 977 F. Supp. 2d, at 1317; but ef. Richmond Medical Center v. Gilmore, 144 F. 3d 326, 330-332 (CA4 1998) (Luttig, J., granting stay).
Regardless, even were we to grant the Attorney General’s views "substantial weight,” we still have to reject his interpretation, for it conflicts with the statutory language discussed supra, at 940. The Attorney General, echoed by the dissents, tries to overcome that language by relying on other language in the statute; in particular, the words "partial birth abortion,” a term ordinarily associated with the D&X procedure, and the words “partially delivers vaginally a living unborn child.” Neb. Rev. Stat. Ann. §28-326(9) (Supp. 1999). But these words cannot help the Attorney General. They are subject to the statute’s further explicit statutory definition, specifying that both terms include “delivering into the vagina a living unborn child, or a substantial portion thereof.” Ibid. When a statute includes an explicit definition, we must follow that definition, even if it varies from that term’s ordinary meaning. Meese v. Keene, 481 U. S. 465, 484-485 (1987) (“It is axiomatic that the statutory definition of the term excludes unstated meanings of that term”); Colautti v. Franklin, 439 U. S., at 392-393, n. 10 (“As a rule, ‘a definition which declares what a term “means” . . . excludes any meaning that is not stated’”); Western Union Telegraph Co. v. Lenroot, 323 U. S. 490, 502 (1945); Fox v. Standard Oil Co. of N. J., 294 U. S. 87, 95-96 (1935) (Cardozo, J.); see also 2A N. Singer, Sutherland on Statutes and Statutory Construction §47.07, p. 152, and n. 10 (5th ed. 1992) (collecting eases). That is to say, the statute, read “as a whole,” post, at 998 [943] (Thomas, J., dissenting), leads the reader to a definition. That definition does not include the Attorney General’s restriction — “the child up to the head.” Its words, “substantial portion,” indicate the contrary.
The Attorney General also points to the Nebraska Legislature’s debates, where the term “partial birth abortion” appeared frequently. But those debates hurt his argument more than they help it. Nebraska’s legislators focused directly upon the meaning of the word “substantial.” One senator asked the bill’s sponsor, “[Y]ou said that as small a portion of the fetus as a foot would constitute a substantial portion in your opinion. Is that correct?.” The sponsoring senator replied, “Yes, I believe that’s correct.” App. 452-453; see also id., at 442-443 (same senator explaining “substantial” would “indicate that more than a little bit has been delivered into the vagina,” i. e., “[e]nough that would allow for the procedure to end up with the killing of the unborn child”); id., at 404 (rejecting amendment to limit law to D&X). The legislature seems to have wanted to avoid more limiting language lest it become too easy to evade the statute’s strictures — a motive that Justice Thomas well explains. Post, at 1001-1003. That goal, however, exacerbates the problem.
The Attorney General, again echoed by the dissents, further argues that the statute “distinguishes between the overall ‘abortion procedure’ itself and the separate ‘procedure’ used to kill the unborn child.” Brief for Petitioners 16-18; post, at 991-992 (opinion of Thomas, J.), 975-976 (opinion of Kennedy, J.). Even assuming that the distinction would help the Attorney General make the D&E/ D&X distinction he seeks, however, we cannot find any language in the statute that supports it. He wants us to read “procedure” in the statute’s last sentence to mean “separate procedure,” i. e., the killing of the fetus, as opposed to a whole procedure, i. e., a D&E or D&X abortion. But the critical word “separate” is missing. And the same [944] word “procedure,” in the same subsection and throughout the statute, is used to refer to an entire abortion procedure. Neb. Rev. Stat. Ann. §§28-326(9), 28—328(1)—(4) (Supp. 1999); cf. Gustafson v. Alloyd Co., 513 U. S. 561, 579 (1995) (“[Ijdentieal words used in different parts of the same act are intended to have the same meaning” (internal quotation marks omitted)).
The dissenters add that the statutory words “partially delivers” can be read to exclude B&E. Post, at 990-991 (opinion of Thomas, J.), 974 (opinion of Kennedy, J.). They say that introduction of, say, a limb or both limbs into the vagina does not involve “delivery.” But obstetric textbooks and even dictionaries routinely use that term to describe any facilitated removal of tissue from the uterus, not only the removal of an intact fetus. E. g., Obstetrics: Normal & Problem Pregnancies, at 388 (describing “delivery” of fetal membranes, placenta, and umbilical cord in the third stage of labor); B. Maloy, Medical Dictionary for Lawyers 221 (3d ed. 1960) (“Also, the removal of a [fetal] part such as the placenta”); 4 Oxford English Dictionary 422 (2d ed. 1989) (to “deliver” means, inter alia, to “dis-burden (a women) of the foetus”); Webster’s Third New International Dictionary (1993) (“ff)]elivery” means “the expulsion or extraction of a fetus and its membranes”). In any event, the statute itself specifies that it applies both to delivering “an intact unborn child” or “a substantial portion thereof.” The dissents cannot explain how introduction of a substantial portion of a fetus into the vagina pursuant to D&X is a “delivery,” while introduction pursuant to D&E is not.
We are aware that adopting the Attorney General’s interpretation might avoid the constitutional problem discussed in this section. But we are “without power to adopt a narrowing construction of a state statute unless such a construction is reasonable and readily apparent.” Boos v. Barry, 485 U. S. 312, 330 (1988); Gooding v. Wilson, 405 U. S. 518, 520-[945]*945521 (1972). For the reasons stated, it is not reasonable to replace the term “substantial portion” with the Attorney General’s phrase “body up to the head.” See Almendarez-Torres v. United States, 523 U. S. 224, 237-239 (1998) (statute must be “genuinely susceptible” to two interpretations).
Finally, the law does not require us to certify the state-law question to the Nebraska Supreme Court. Of course, we lack any authoritative state-court construction. But “we have never held that a federal litigant must await a state-court construction or the development of an established practice before bringing the federal suit.” City of Lakewood v. Plain Dealer Publishing Co., 486 U. S. 750, 770, n. 11 (1988). The Attorney General did not seek a narrowing interpretation from the Nebraska Supreme Court nor did he ask the federal courts to certify the interpretive question. See Brief for State Appellants in Nos. 98-3245 and 98-3300 (CA8); cf. Arizonans for Official English v. Arizona, 520 U. S. 43 (1997). Even if we were inclined to certify the question now, we cannot do so. Certification of a question (or abstention) is appropriate only where the statute is “fairly susceptible” to a narrowing construction, see Houston v. Hill, 482 U. S. 451, 468-471 (1987). We believe it is not. Moreover, the Nebraska Supreme Court grants certification only if the certified question is “determinative of the cause.” Neb. Rev. Stat. §24-219 (1995); see also Houston v. Hill, supra, at 471 (“It would be manifestly inappropriate to certify a question in a ease where . . . there is no uncertain question of state law whose resolution might affect the pending federal claim”). Here, it would not be determinative, in light of the discussion in Part II-A, supra.
In sum, using this law some present prosecutors and future Attorneys General may choose to pursue physicians who use D&E procedures, the most commonly used method for performing previability second trimester abortions. All those who perform abortion procedures using that method must fear prosecution, conviction, and imprisonment. The [946] result is an undue burden upon a woman’s right to make an abortion decision. We must consequently find the statute unconstitutional.
The judgment of the Court of Appeals is
Affirmed.