Marjorie Griffin and Sandra McWhorter appeal the judgment of the district court denying their claims of racial and sexual discrimination brought under Title VII of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000e-2(a)(l)-(2) (1982), and 42 U.S.C. §§ 1981, 1983 (1982). The appellants’ claims arise out of their termination from the Omaha Police Department recruit class • of 1981 for failure to meet, the minimum firearms proficiency standards. Griffin and McWhorter, both black women, contend that they received inadequate firearms training, qualitatively and quantitatively inferior to the training provided other recruits; that they were terminated from the recruit program for failure to satisfy firearms qualification standards while a similarly situated white male was retained in the program; and, that they were subjected to a racially discriminatory work environment. In addition to their claims of disparate treatment and discriminatory work environment under Title VII, and intentional discrimination under sections 1981 and 1983, the appellants contend that the Police Department’s facially neutral firearms qualification requirements disproportionately exclude blacks and women and thereby violate the disparate impact provisions of Title VII. The appellants argue on appeal that the district court’s findings, that they were not provided inferior training nor terminated from the recruit program because of their race or sex, nor subjected to a discriminatory environment, are clearly erroneous, and they seek reinstatement into the police department. On review, we are left with the firm impression that a mistake has been made and that certain findings of the district court are clearly erroneous. We reverse the judgment and remand the cause for further consideration consistent with this opinion.

The Omaha Police Department’s (OPD) 1981 recruit class consisted of 34 individuals: fourteen white males, twelve black males, three hispanic males, two white females and three black females. The 1981 recruit class had the greatest number of black officers in the OPD’s history. Twenty-five members of the 1981 recruit class graduated; all nine recruits terminated were black, and included all three black females.1

In October 1980, approximately one year before the appellants joined the OPD as recruits, the City of Omaha (City) entered into a consent decree which required an increase in black employment in the OPD.2 The district court found that the hiring requirements of the consent degree had caused considerable racial tension within the OPD ranks. The district court further found that rumors had circulated among officers and recruits that black recruits would be “washed out” of the 1981 recruit program.3 Additionally, a number of black *623recruits had been experiencing academic difficulties and had expressed concern to members of the Brotherhood of Midwest Guardians, an organization of black police officers, that black recruits were being ignored in class and were receiving training inferior to that provided white recruits. The district court found that the defendants, upon learning of the black recruits’ concerns, took steps to improve the work atmosphere and promote more positive feelings within the ranks. In addition, the OPD administration, along with the City’s Affirmative Action Officer and the Midwest Guardians, began actively to monitor the situation. The Public Safety Director, in charge of the OPD, personally investigated the matter and discussed it with the Midwest Guardians and the OPD training personnel. Several minority recruits testified at trial that they did not perceive nor were they subjected to racial hostility from the training staff.

The bulk of the evidence presented at trial concerned two issues: what were the OPD firearms proficiency standards, and did the appellants meet those standards; and, whether the training provided the appellants and other recruits was adequate to qualify with firearms.

Before the 1980 consent decree, the OPD had no written firearms standards. However, it regularly and uniformly had terminated any recruit who failed to achieve a score of 65 on any of six pre-final qualification shoots, with one make-up shoot allowed to achieve a passing score. A written policy was drawn up in December 1980 which set the minimum record score — the original shoot or the make-up shoot — for any pre-final qualification shoot at 65, and permitted only one make-up shoot for any one pre-final qualification shoot failed. The written policy also required a recruit to maintain an average of 70 or better in the six pre-final qualification shoots. Additionally, the policy provided that any recruit who failed to score 70 or better on any qualification shoot would be given necessary corrective assistance. If a recruit’s record score fell below 65, the recruit would be recommended for termination; if the recruit’s average score for the six prefinal qualification shoots was below 70, the recruit also would be recommended for termination. Lastly, the policy provided that a recruit’s score on a final qualification shoot would be averaged with the pre-final scores to determine the recruit’s “shooting score”; no make-ups would be allowed for this final qualification shoot. The policy did not set any specific score that must be achieved on the final shoot. Addendum to Appellants Opening Brief at 40.

The recruits underwent firearms training and practice prior to the pre-qualification shoots. Neither Griffin nor McWhorter had used a firearm before. After the first pre-final qualification shoot, approximately seven weeks into the training, no regularly scheduled practice sessions were provided. The appellants testified that the training instructors failed to criticize their technique or provide instruction during the training and practice phase despite their evident problems with firearms. They further testified that the training officers did not assist them properly to correct their errors during the first five pre-qualification shoots, despite the fact that Griffin failed one shoot and McWhorter failed three of the first five shoots.

On November 5, 1981, Griffin, McWhorter and Rosalyn Cotton,4 the third black female, failed the sixth pre-final qualification shoot. For the first time, a white male, William Dussetschleger, failed as well. The district court found that all four recruits were thereafter recommended for termination. The three black female recruits complained to the Midwest Guardi*624ans that the firearms training they had received was inadequate; the Guardians in turn brought these complaints to the attention of the OPD administration. After considering these complaints, the OPD administration revised the firearms qualification policy and permitted all four recruits to remain in the recruit class. Under the revised standard, recruits would be certified upon completion of the entire training if they achieved a minimum average of 70. Addendum to Appellant’s Opening Brief at 43. The administration ordered the training officers to give the four recruits additional assistance after hours to prepare them for the final qualification shoot. The Acting Chief of Police also ordered the training officers to submit to him written reports indicating the nature of the recruits’ problems and their progress.

When it became apparent that Griffin, McWhorter and Rosalyn Cotton were having difficulty with firearms training, Maggie Heaston, the City Affirmative Action Officer, suggested to Public Safety Director Joe Friend a number of additional methods to remedy the inadequate firearms training allegedly provided the black women recruits.5 Friend rejected these suggestions without discussion.

Consistent with Friend’s orders, the four recruits were provided additional training. Recruit Dussetschleger was assigned to the chief firearms training officer, Officer Benak; Griffin and Cotton were assigned to Officer Pekula, and McWhorter was assigned to Officer Schlotman. Louis Dirks, Training Officer at the Law Enforcement Training Center at Grand Island, Nebraska,- testified as appellant’s firearms expert. According to Dirks, who had reviewed the training officers’ reports to the Acting Chief of Police, Officer Benak’s reports indicate that he made an early diagnosis of Dussetschleger’s problems and devised a remedy which proved successful. Dirks complimented the quality and specificity of Benak’s diagnosis and suggested remedies. Dussetschleger passed the final qualification shoot with a score of 87.2, which gave him an overall average score for the qualification shoots of 76.1.

On the other hand, Dirks criticized the reports filed by Officers Pekula and Schlotman. He testified that they did not specifically diagnose the recruits’ individual problems, but merely reiterated “basics,” and suggested inappropriate remedies. The officers admitted at trial that their reporting was “negligent,” but both provided an oral account of their diagnoses of the recruits’ problems. Griffin scored a 64.8 on . the final qualification shoot, which brought her overall average to 73.1. McWhorter scored a 47.6 on the final qualification shoot, which brought her overall average to 66.3. Cotton also failed to score a 65 on the final shoot, and her average was below 70. Griffin, McWhorter and Cotton, all of whom had satisfactorily completed all other training activities and testing required of police recruits, were terminated. Dussetschleger graduated from the recruit program.

Testimony from expert witnesses called by both parties established that termination for failure to meet firearms standards is rare. Dirks testified that both Griffin and McWhorter came to the Grand Island training center after they had been terminated from the OPD. He testified that he was able to diagnose their problems and qualified both within a weekend.

The district court found that Griffin and McWhorter had not been given satisfactory instruction or guidance during the early stages of firearms training, but that the training given other recruits was just as unsatisfactory. The court concluded that the firearms training provided the appellants was not quantitatively or qualitatively different from that provided other recruits because of the appellants’ race or sex. The district court also concluded that the appellants were not terminated on the *625basis of race or sex. The court found that while they had failed to meet the OPD firearms requirement, Dussetschleger, a white male, had improved steadily during practice sessions and met the minimum proficiency standards. Thus, the district court concluded, based on this evidence, the appellants had failed to make out a prima facie case of disparate treatment.

The district court also concluded that the appellants had failed to make out a case of disparate impact. The district court recognized that hand strength is a factor in marksmanship, and particularly important for handgun accuracy at longer distances, but concluded that “beyond this testimony, no other evidence was offered to prove that women as a group, are statistically inferior to men in regard to hand strength or general ability to handle firearms.” Griffin v. City of Omaha, CV-83-0-186, slip op. at 20 (D.Neb. Mar. 4, 1985). The court concluded that the evidence was not sufficient to demonstrate that females have a greater incidence of error in training in the use of firearms.

The district court also concluded that the appellants had failed to prove that racism sufficiently pervaded the work environment to constitute a violation of Title VII. Finally, the district court concluded that the appellants had failed to prove a violation of sections 1981 and 1983. The court reasoned that the legal analysis under sections 1981 and 1983 is essentially the same as that which is applied to disparate treatment claims under Title VII.I. ****6

Griffin and McWhorter argue on appeal that the district court erred in rejecting their disparate treatment and disparate impact claims and in finding that the City did not discriminate against them with respect to either their firearms training or their discharge.

I.

Proof of disparate treatment under Title VII follows a well-established pattern. The plaintiff initially must establish a prima facie case by proving facts sufficient to give rise to an inference of discrimination. If the plaintiff succeeds in establishing a prima facie case, the burden shifts to the defendant to articulate a legitimate, non-discriminatory reason for the allegedly impermissible employment decision. If these initial burdens are met, the plaintiff is then afforded an opportunity to show that the employer’s justifications for the decision are mere pretexts, concealing improper motivation. McDonnell Douglas Cory. v. Green, 411 U.S. 792, 802-805, 93 S.Ct. 1817, 1824-1825, 36 L.Ed.2d 668 (1973); see also Patterson v. Masem, 774 F.2d 251, 254 (8th Cir. 1985). The district court, following this analytic pattern, held that Griffin and McWhorter had failed to make out a prima facie case that the City had impermissibly discriminated against them in either their firearms training or their termination for failure to meet the minimum firearms proficiency standards. Our review of the district court’s findings, however, need not follow this strict sequence. We may look directly to the ultimate factual issue: whether the City intentionally discriminated against the appellants. United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 715, 103 S.Ct. 1478, 1481, 75 L.Ed.2d 403 (1983); see also Patterson, 774 F.2d at 254; Craft v. Metromedia, 766 F.2d 1205, 1211 (8th Cir. 1985).

Appellate review of a district court’s factual findings is limited, however, by the “clearly erroneous” standard. Fed. R.Civ.P. 52(a); see Anderson v. City of Bessemer City, — U.S. —, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985). We may not reverse the district court’s findings that the appellants were not subject to impermissible discrimination unless our review of the record leaves us with the “definite *626and firm conviction that a mistake has been made.” Anderson, 105 S.Ct. at 1511 (quoting United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 542, 92 L.Ed. 746 (1948)). Reviewing courts must refrain from duplicating the function of the trial court by attempting to decide factual issues de novo after an independent review of the record. Id. at 1511— 12. The Supreme Court made clear, however, that a district court’s findings are not entirely insulated from appellate review, even if they are based on witness credibility. Where physical, documentary or other forms of objective evidence contradict a witness’s story, or that story is so internally inconsistent or implausible th'at a reasonable factfinder would not credit it, a reviewing court may find clear error even “in a finding purportedly based on a credibility determination.” Id. at 1513. The burden is on the objecting party to demonstrate clear error in factual findings, and the evidence must be construed in the light most favorable to the party who prevailed at trial. Craft, 766 F.2d at 1212.

From our review of the record, we are left with the firm impression that the district court mistakenly concluded that neither Griffin nor McWhorter was subject to intentional discrimination on the basis of race. We are satisfied that Griffin met the applicable firearms standards. Further, with respect to McWhorter, we believe that the district court failed to consider important evidence and testimony, some of which contradicts evidence and testimony upon which the court premised its findings.

We turn first to the evidence with respect to the firearms qualification standards applied to the recruit class. The district court found that the OPD standards required a recruit’s disqualification if she scored below 65 on any shoot — pre-final or final qualification — or failed to maintain an overall average score of 70. The district court mechanically applied these standards to the appellants’ scores and determined that they had failed to meet the minimum standards, while the white male recruit who experienced problems with firearms training had satisfied them.

We believe the court clearly erred in finding that Griffin and McWhorter failed to satisfy the official OPD firearms standards. Public Safety Director Friend, who had ultimate responsibility for terminating the appellants, testified that he employed a subjective “deterioration" standard, not the official OPD standard. Friend repeatedly admitted that he terminated the appellants because their scores deteriorated and fell below 65.7 Thus, according to the testimony of the official responsible for the decision, the appellants were not discharged for failure to meet the announced OPD firearms standards, but for failure to meet a new, subjective standard. The district court’s finding that the appellants were terminated because they failed to comply with the established OPD firearms proficiency standards therefore is clearly erroneous.8

*627Moreover, we believe the district court erred in finding the official OPD firearms standard required a recruit to score 65 or above on the final qualification shoot. The first written standards, introduced in 1980, required that a recruit maintain an average of 70 or better for the six pre-final qualification shoots, though a recruit with a record score on any pre-final shoot— original or make-up — under 65 would be recommended for termination. With regard to the final qualification shoot, the written policy stated only that the score for this shoot — for which no make-up would be allowed — would be averaged with the prefinal qualification scores to determine the recruit’s shooting score. The written policy does not provide that a recruit must score a 65 or better on the final shoot.

These standards were revised on November 6, 1981, in response to the appellants’ complaints of inadequate training, to permit recruits to be certified if they had achieved an average of 70 upon completion of the training. This policy revision was explicitly acknowledged in a letter from the Assistant City Attorney to the Nebraska Equal Opportunity Commission responding to the appellants’ charges filed with that administrative body. The district court did not discuss the written firearms policy, nor take notice on the record of the City’s admission in its filing with the NEOC. This admission contradicts and discredits the testimony at trial suggesting that the OPD official policy required a recruit to score a 65 on the final qualification shoot. Cf. Anderson, — U.S. at-, 105 S.Ct. at 1512 (appellate court finding of clear error may be based on objective or documentary evidence which contradicts witnesses’ testimony).

This evidence established that the official OPD firearms policy required a recruit only to maintain a 70 average after all qualification shoots. The City acknowledged, and the district court found, that Griffin satisfied this requirement; the testimony demonstrated that Griffin maintained an overall score of 73.1. Therefore, had the Public Safety Director applied the official OPD firearms standards to Griffin, he would necessarily have concluded that she passed the firearms qualification.

Therefore, we conclude that the district court was clearly erroneous in finding that Griffin was dismissed for failing to meet the OPD firearms standards. We therefore reverse the judgment of the district court with respect to appellant Griffin, and remand with directions that it enter a judgment in favor of Griffin and determine the remedy that will make her whole.

McWhorter, on the other hand, manifestly did not satisfy the firearms standards, announced or applied. This, she claims, resulted from inferior, discriminatory firearms training. The district court found that appellants’ training, though perhaps poor, was not materially different from that provided the other recruits. We believe that there is testimony in the record, none of which was the subject of specific findings by the district court nor explicitly referred to in its memorandum opinion, which appears to strongly contradict evidence and testimony upon which it relied in concluding that the appellants had not received inferior firearms training. We recognize, of course, that a district court’s failure to explicitly account for evidence favorable to the appellant does not render its decisions clearly erroneous, nor even suspect. Grebin v. Sioux Falls Independent School District No. 49-5, 779 F.2d 18, 19 (8th Cir. 1985); Talley v. United States Postal Service, 720 F.2d 505, 507 (8th Cir. 1983), cert. denied, 466 U.S. 952, 104 S.Ct. 2155, 80 L.Ed.2d 541 (1984). The trial *628court need not make specific findings with respect to all the evidence presented, nor even refer to all the evidence introduced, particularly when the trial is lengthy and complex, as was this one. See Grebin, 779 F.2d at 19; Talley, 720 F.2d at 507. However, Rule 52 does not free the district court of the burden of assuring the appellate court, through findings of fact or references in its memorandum opinion, that it has considered strongly conflicting evidence and “come to grips with apparently irreconcilable conflicts.” Tate v. Weyerhauser Co., 723 F.2d 598, 605 (8th Cir. 1983) (quoting EEOC v. Federal Reserve Bank of Richmond, 698 F.2d 633, 640 (4th Cir. 1983), rev’d on other grounds sub nom. Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 104 S.Ct. 2794, 81 L.Ed.2d 718 (1984)), cert. denied, — U.S. —, 105 S.Ct. 160, 83 L.Ed.2d 97 (1984). A district court’s findings are clearly erroneous “to the extent that they fail to recognize [important] incidents and reject or fail to draw the inferences which we have found inescapable from the record.” Alexander v. National Farmers Organization, 687 F.2d 1173, 1203 (8th Cir. 1982), cert. denied, 461 U.S. 937, 103 S.Ct. 2108, 77 L.Ed.2d 313 (1983).

Specifically, we believe the district court should have accounted for the uncontradicted testimony of virtually every firearms expert who testified that it is extremely unusual that an individual, male or female, cannot be trained to qualify in firearms, and the testimony of Dirks that both appellants were qualified at the state police training center within one weekend.

The City’s firearms expert, Robert Monroe, a retired FBI agent, stated that it would be an unusual occurrence to be unable to train a person to qualify with firearms, although he acknowledged that it does happen. Although he was unable to recall from the thousands of individuals he has trained how many he was unable to qualify, he testified that he never had 10 percent of a class that failed to qualify. He further testified that he never had trained a class where he could not train 60 percent of the females and 100 percent of the black females. Monroe admitted that he had never had a particular class in which three women had failed. Captain (then Lieutenant) Sorys, in charge of the OPD recruit training program during the relevant period, testified that of 225 to 230 recruits that had been at the academy since he has been associated with the Training Division, only 4 had failed to pass firearms qualification.9 Dirks testified that of 1100 recruits, he had failed to qualify only one in firearms.

Where the discharge of three black females from one recruit class for failure to qualify with firearms was the primary issue before the court, we believe that this testimony, passed over by the court, is of extreme significance.

We believe, therefore, that the appropriate course is to remand this cause to the district court with respect to recruit McWhorter. The district court should reconsider its decision in light of the issues and evidence which we have discussed above. As we have noted, “appellants should be given the full benefit of their proof without tightly compartmentalizing the various factual components and wiping the slate clear after scrutinizing each.” Alexander, 687 F.2d at 1207-08 (quoting Continental Ore Co. v. Union Carbide Carbon Co., 370 U.S. 690, 699, 82 S.Ct. 1404, 1410, 8 L.Ed.2d 777 (1962)).

The testimony made abundantly clear the need for firearms standards for police officers. Our concern, however, is with application of the standards and the inconsistencies apparent in the record of this case, particularly when the failure to qualify seems to be such a rare occurrence, and where it fell on three black women who made strong arguments of inadequate training.

We reverse the judgment of the district court and remand for determination of a *629proper remedy for Griffin, and for further consideration of McWhorter’s claims consistent with this opinion.