After finding that the Mexican-American plaintiffs had proved a prima facie case of discrimination in the use of certain hiring tests for the position of deputy sheriff by Los Angeles County, the district court went on to find that validation of the challenged tests was sufficient to rebut any charge of a violation of federal law. Consequently the court entered judgment dismissing the action. Plaintiffs appeal. We reverse in part and affirm in part.

This class action challenged the defendants’ use of two written examinations plus the minimum height requirements of the sheriff’s department as violations of Title VII of the Civil Rights Act of 1964, as amended (42 U.S.C. §§ 2000e, et seq.), and the Civil Rights Act of 1870 ( 42 U.S.C. § 1981 ) .

The trial court considered two written verbal and cognitive ability tests. Statistics demonstrated that the ET4-1 test, administered between 1973 and 1975, had a substantial adverse impact on Mexican-Americans, requiring validation of the test. We will examine the sufficiency of that validation in due course.

The district court did not reach the validation issue for the second test (CPS Form 51X) employed by the sheriff. Based on its finding that 53% of the Mexican-Americans failed the test compared to approximately 50% of the whites, the court concluded that use of that test did not adversely affect Mexican-Americans. The court’s finding on impact is clearly erroneous. Stipulated statistics in the record reveal that 33% of the Mexican-Americans who took the 51X test failed while only 13% of the whites failed. This difference is sufficient to establish a prima facie case of discrimination under Title VII. On remand, the defendants will have an opportunity to attempt to validate the CPS Form 51X examination, if they wish to continue to use it.

I. Validation

Once the plaintiff has established a prima facie case of discrimination, the burden shifts to the employer to justify the challenged selection device as a business necessity by showing that it is significantly job-related. Dothard v. Rawlinson, 433 U.S. 321 , 97 S.Ct. 2720 , 53 L.Ed.2d 786 (1977); Blake v. City of Los Angeles, 595 F.2d 1367, 1378 (9th Cir. 1979). This process is generally referred to as “validation.” Id. at 1377 . The validation burden compels the employer to show that the preemployment selection device has “a manifest relationship to the employment in question.” Griggs v. Duke Power Co., 401 U.S. 424, 432 , 91 S.Ct. 849, 854 , 28 L.Ed.2d 158 (1971). The Supreme Court elaborated on the legal standard governing this inquiry as follows:

The validation process thus involves three distinct steps. The employer must first specify the particular trait or characteristic which the selection device is being used to identify or measure. The employer must then determine that that particular trait or characteristic is an important element of work behavior. Finally, the employer must demonstrate by “professionally acceptable methods” that the selection device is “predictive of or significantly correlated” with the element of work behavior identified in the second step. Id.

We turn now to the defendants’ attempted validation of the ET4-1 test. The sheriff’s department uses this test to determine whether the job applicant is likely to succeed in the required academic training program for the position of deputy sheriff.

The plaintiffs assert that it is not permissible under Title VII to test for predicted performance at the Sheriff’s Academy; they would limit the sheriff to testing directly for predicted performance on the job. The law is somewhat ambiguous on this issue. This court has recently suggested by way of a footnote that Title VII does not permit correlation between a selection device and academy performance to substitute for validation of the device vis-a-vis actual job performance. Blake, 595 F.2d at 1382 -83 n.17. The court there required the defendant police department to demonstrate a significant relationship between the challenged physical abilities test and important aspects of actual job performance as a police officer. The court recognized that one way the defendant might satisfy the requirement would be to show that the selection test correlates with measures of success in training and that those measures in turn are significantly related to job performance.

Prior to Blake , the Supreme Court had addressed the question of validation of a preemployment selection test by demonstrating relationship with an academic training program. Washington v. Davis, 426 U.S. 229, 249-52 , 96 S.Ct. 2040 , 48 L.Ed.2d 597 (1976). The Court there approved the defendant police department’s use of a verbal abilities test to determine whether applicants possessed the minimum skills necessary to understand and succeed in the academic portion of police officer training. In the circumstances before it, the Court agreed with the “sensible construction of the job-relatedness requirement” advanced by the three-judge district court to the effect that “training-program validation may itself be sufficient” to validate a selection device which impacts heavily on a protected group. Id. at 250-51 , 96 S.Ct. at 2052-2053 . (Emphasis added.)

The Supreme Court later summarily affirmed a case which found the validation of national teacher certification examinations against academic teacher training programs, rather than against job performance, sufficient to satisfy the job-relatedness requirement of Title VII. National Education Ass’n v. South Carolina, 434 U.S. 1026 , 98 S.Ct. 756 , 54 L.Ed.2d 775 (1978), aff’g United States v. South Carolina, 445 F.Supp. 1094 (D.S.C. 1977).

In neither Washington v. Davis nor National Education Ass’n does it appear that the Supreme Court considered a potential danger this circuit identified in Blake :

The facts in the cases before the Supreme Court suggest that this danger was minimal or nonexistent. The preemployment test in Washington v. Davis was “ ‘designed to test verbal ability, vocabulary, reading and comprehension,’ ” 426 U.S. at 235 , 96 S.Ct. at 2045 , and was used by the police department to assure that candidates possessed “some minimal verbal and communicative skill” necessary to enter the police training program. Id. at 250 , 96 S.Ct. at 2052 . The Supreme Court not unjustifiably assumed that those same skills were relevant to actual job performance. In National Education Association, the validation study demonstrated that the certification tests correlated with the content of courses at teacher training institutions in South Carolina. In that situation, there is little danger that both the qualifying test and the academic program would be manipulated to defeat the purposes of Title VII. Academic training, which there precedes the testing, is not in the hands of the immediate employer.

However, the instant case does not demonstrate similar safeguards. Both the selection procedures and the training program are under the control of the employer. And the ET4 — 1 aptitude test is utilized to measure, not just minimum communication skills, but also cognitive and writing skills.

Thus, we hold that the skill, characteristic or ability which the ET4-1 test purports to measure must be important to the requirements of the job. The sheriff’s department may show the requisite relationship by demonstrating that the test correlates significantly with important elements of academy performance and that those elements of academy performance are important to actual job performance.

As specified earlier, the second stage in the validation process requires the defendant to show that the skill or characteristic which the challenged selection device is used to identify or measure is an important element of work behavior comprising or relevant to the job for which applicants are being screened. Albemarle Paper Co., 422 U.S. at 431 , 95 S.Ct. at 2378 . A specific skill or characteristic is an important element of job or training program performance when a significant portion of the employee or candidate time and energy on the job or in the training program calls for utilization of that skill or characteristic or when its absence would significantly impair the employee’s or candidate’s effectiveness.

In the instant case, the sheriff’s department has shown only that the ET4-1 test correlates with performance on the academic portion of academy training. Specifically, the ET4-1 test measures the applicant’s reading, writing and reasoning skills. These skills are considered essential to acceptable performance in such courses as constitutional law, criminal law and procedure, and juvenile law and procedure, which are taught at the sheriff’s academy.

The Los Angeles County Sheriff’s Department is a large police agency engaged in many aspects of police work. To prepare deputies for the increasingly complex requirements of police work, the department has placed heavy emphasis on law-related academy courses as well as training required in technical areas like telecommunications. We believe that the emphasis is justified in light of the professional and constitutional demands that affect police work.

Given that a significant amount of a new deputy’s training in legal areas properly occurs in the classroom, the sheriff has established that some level of verbal and cognitive skills is essential to successful completion of the academy program. What the appropriate level is, however, remains to be demonstrated.

We turn now to a discussion of the final stage of the validation process. The question here is whether the sheriff has demonstrated, by acceptable means, a significant correlation between the ET4-1 test and the important elements of job and academy performance.

The requirements at this stage are twofold: (1) the validation study must use a professionally acceptable method of showing correlation, and (2) the correlation must be “predictive” or “significant”. Albemarle Paper Co., 422 U.S. at 431 , 95 S.Ct. at 2378 .

A number of validation techniques are professionally acceptable. See Washington v. Davis, 426 U.S. at 247 n.13, 96 S.Ct. at 2051 (describing three professional methods of validation). The Equal Employment Opportunity Commission (“EEOC”) has published guidelines which systematically outline minimum standards for professional validation studies. 29 C.F.R. § 1607 (1975). EEOC guidelines are “entitled to great deference” from courts faced with deciding whether particular employment selection devices comply with Title VII. Griggs v. Duke Power Co., 401 U.S. at 434 , 91 S.Ct. at 855 . See also Washington v. Davis, supra, 426 U.S. at 247 n.13, 96 S.Ct. at 2051 . Although compliance with the EEOC guidelines is not mandatory, satisfactory validation of a preemployment selection device may be more difficult to establish when the guidelines have not been followed. United States v. City of Chicago, 573 F.2d 416, 426-27 (7th Cir. 1978). In addition, methodological or other defects in validation studies diminish the probative value of the obtained results. Blake v. City of Los Angeles, 595 F.2d at 1380-81 .

Besides being professionally acceptable, the validation study must demonstrate a significant relation between the challenged selection device or criteria and the important elements of the job or training program, not merely some “rational basis” for the challenged practice. Washington v. Davis, 426 U.S. at 247 , 96 S.Ct. at 2051 . However, the employer need not establish a perfect positive correlation between the selection criteria and the important elements of work.

The ET4-1 validation study offered in this case produced a significant correlation coefficient (.60) between the test and academic performance as measured by scores on written examinations given at the end of training. However, plaintiffs enumerate several defects in the sheriff’s study which they argue are enough to render the established correlation meaningless.

Plaintiffs first attack the sample utilized in the validation study as unrepresentative of the applicant pool because only persons who received passing scores on the ET4-1 were included in the sample. Plaintiffs note that, as a result, no one really knows whether some who failed the test would nonetheless have successfully completed the training program. In support of this criticism of the validation study, plaintiffs cite EEOC guidelines which state that “[w]here a validity study is conducted in which tests are administered to applicants, with criterion data collected later, the sample of subjects must be representative of the normal or typical candidate group for the job or jobs in question.” 29 C.F.R. § 1607.5 (b)(1) (1975). Plaintiffs also cite the following language from Blake v. City of Los Angeles, supra, as supportive of their challenge:

We agree with the Blake construction of the cited EEOC guidelines. See also United States v. Georgia Pacific Co., 474 F.2d 906, 915-16 (5th Cir. 1973). And, as we earlier noted, noncompliance with the EEOC guidelines diminishes the probative value of the defendants’ validation study. But it is not necessarily fatal.

The sheriff’s department demonstrated a very high correlation between scores on the ET4-1 test and academic averages at the academy. Plaintiffs point to no data which are inconsistent with defendants’ validation study. Nor do they cite serious flaws in the study’s data-gathering or handling which might warrant the suspicion that the obtained correlation is spurious or erroneous.

In contrast, the Blake plaintiffs raised a host of methodological criticisms of the defendants’ validation studies, including ambiguous definitions of the crucial variable, potential coder bias, and lack of control for experience. In addition, they alleged that other research as well as one of the defendants’ own studies resulted in some data directly contrary to the defendants’ position. Plaintiffs’ showing in that case clearly entitled them to reversal of the summary judgment entered against them.

Under the circumstances here, however, the demonstrated correlation is sufficiently strong to support the inference drawn by the district court after trial that persons excluded from the academy, and thus from the sample, because they failed the ET4-1 entrance exam would not have succeeded in the academic training program. We perceive no reason at this point to require the sheriff’s department to hire and train a sample of failing applicants, with the attendant expense to the county and potential unfairness to the candidates so hired, in order to provide statisticians with more certain results.

The plaintiffs also contend that the sheriff’s department failed to validate the passing score cutoff for ET4-1. The EEOC guidelines do not require separate validation of passing scores. They do require that the cutoff be “reasonable and consistent with normal expectations of proficiency within the work force or group on which the study was conducted.” 29 C.F.R. § 1607.6 (1975). The sheriff’s department complied with this requirement. The evidence indicated that those candidates who barely passed the ET4-1 examination performed correspondingly in the academic classes at the academy. A lower passing score would have resulted in the admission of a number of applicants who would not have been able to perform acceptably in the academy.

Plaintiffs have presented no evidence indicating that the cutoff score was set arbitrarily. Nor was it so high as to maximize adverse impact on Mexican-Americans. While a lower passing score would have increased their rate of passing, 55 percent of all Mexican-Americans taking the test between 1971 and 1975 were in fact able to achieve the passing score or higher. Thus, the cutoff level appears to us to have been reasonable and consistent with normal expectations of performance in the academic training program.

Plaintiffs next complain that the ET4-1 test was not shown to correlate with other important elements of sheriff’s academy performance such as leadership ability, ability to get along with other trainees, and practical application of academic knowledge. The plaintiffs misconstrue an employer’s obligation under Title VII. The sheriff’s department must establish that the challenged selection device is a strong predictor of or is significantly correlated to important elements of work or academy performance. It is not compelled to demonstrate a significant correlation to all elements of work or training. Once the sher iff’s department showed that success on the ET4-1 examination correlated significantly with academic performance, then it has satisfied that part of its burden.

Plaintiffs finally attempt to discredit the defendants’ validation study by pointing out the lack of analysis of the ET4-1 test for “practical significance.” We fail to understand how this criticism differs significantly from the last one considered and thus find it to be equally without merit.

Thus, as to the ET4-1 examination, we remand solely to allow defendants the opportunity to demonstrate the second link in the validation chain. That is, they must now show that the level of academic training given to new recruits manifestly relates to the knowledge that deputy sheriffs must possess to perform the job successfully. We assume that that level lies somewhere above the understanding of the ordinary layperson yet well below that required of an effective legal advocate or criminal law scholar. But an assumption is not an acceptable substitute for evidence of validity. See 29 C.F.R. § 1607.8 (1975). As part of the requisite showing on remand, defendants must also establish that the after-training examinations, with which the ET4-1 test has already been shown to correlate, fairly measure the trainees’ mastery of the subject matter taught in the academic program.

We turn now to the height requirements for deputy sheriffs which the defendants concede had a significant adverse impact upon Mexican-American applicants of both sexes. Although the district court correctly treated the impact as sufficient to make out plaintiffs’ prima facie case, and also noted that the mere existence of the minimum height rule deterred substantial numbers of otherwise potentially qualified Mexican-Americans from applying, the court concluded ultimately that the height requirements were sufficiently job-related to withstand all challenges, and denied relief. In making this ruling in 1977, the court did not have the benefit of this court’s instructive decision in Blake v. City of Los Angeles, supra, which reversed a summary judgment for the Los Angeles Police Department and remanded for trial the question of validating a minimum height requirement that disproportionately excluded women from the police force.

In the case at bar, the district court did take testimony on the validation of the height requirements, and found that they passed scrutiny under Title VII standards. However, the court failed to apply the rigorous standards of job-relatedness thát Title VII demands under the rulings of the Supreme Court and of this circuit.

Recognizing that no professional validation studies had been attempted, the district court further opined that no professionally accepted validation method could demonstrate that minimum height requirements are job-related. Then, relying upon a concurring opinion in Dothard v. Rawlinson, 433 U.S. 321, 337 , 97 S.Ct. 2720, 2730 , 53 L.Ed.2d 786 (1977) (a prison guard case) , the court accepted the conventional belief of police officers, which the court found to be based upon work experience and common sense, that “size is important in discouraging violence.”

As noted earlier, validation of a selection device under Title VII consists of three steps. The first requires the employer to articulate the skill, characteristic or ability which the device is used to identify or measure. Here, the sheriff’s department believed that large body size or stature was related to the ability of deputy sheriffs to command respect and to apply physical restraints when necessary in the control aspects of the job.

In the second step, the employer must demonstrate that the abilities identified in step one are important to job performance. In determining whether deputy sheriffs spend a significant portion of their time engaging in control functions, or whether some inability to “command respect” from or to physically control others would detract from effective job performance, the sheriff’s department cannot rely on generalizations about police “experience.” As the department’s own counsel argued in support of educational testing, police work has become very complex, demanding a multitude of skills and a depth of background.

One generalization on which the sheriff would have us rely is that all police officers expend a significant portion of their time and effort in confrontation and control conditions. This is problematical. No study proved the generalization to be a fact.

The role of a police officer in the community is not the same as that of a prison guard. Although police officers may be compelled to appear credible in crowd dispersal, breaking up fights, and similar functions, officers must also be able to communicate with their constituency, relate to the person on the street, and apply a variety of skills in a variety of exigencies.

While we are inclined to accept the notion that the appearance of strength and the ability to command respect or obtain psychological advantage in a confrontation situation are important elements of some police work, the record does not support acceptance of the generalization that physical size is important in all police positions. Police work may have many functions. One might deal with juvenile or domestic problems, while another specializes in foot or mounted patrol work. A third might engage solely in undercover or investigatory activities. Still another might be limited to education or community relations work.

Even were we to assume that the “control function” identified by the sheriff’s department is important to the effective performance of every deputy, the department has not demonstrated by professionally acceptable methods that the height requirements were predictive of or significantly correlated to that element of job performance.

In the trial of this case, the sheriff and a physical training instructor each testified that height was “extremely important” in the control functions of a deputy sheriff, particularly in commanding respect and in applying physical restraint. The record, however, is devoid of any basis for these opinions. Alone, they are clearly insufficient to satisfy the requirements of validation as set out in Albemarle Paper Co., Dothard, and Blake .

Besides failing to offer any validation evidence demonstrating the asserted relationship between height and the ability to apply physical restraint, the sheriff’s department has not explained why that ability cannot be measured directly. The relationship with height is neither clear nor direct, and use of that criterion has a devastating effect on Mexican-Americans, many of whom may well be capable of subduing larger persons. We agree with the suggestion concurred in by a majority of the Supreme Court in Dothard v. Rawlinson, supra :

The second proffered justification for the height requirement, that it gives the deputy a psychological advantage in confrontation situations, is equally unconvincing. We note that Los Angeles County is a metropolitan area with a large Mexican-American community. The county’s demography makes it inevitable that a substantial amount of deputy-sheriff time will be spent in that community. The department’s argument in effect ignores the respect likely to be given by that community to Mexican-American deputies, however, small in stature, by virtue of their shared language and culture.

Even in a direct confrontation situation, the sheriff’s department has presented no data supporting its contention that assailants are more likely to attack deputies shorter than 67 inches (63 inches for women) than those who are taller. That a hypothesis positing such a relationship may not seem unreasonable is insufficient to satisfy the employer’s burden of proving job-relatedness. See Washington v. Davis, 426 U.S. at 246-47 , 96 S.Ct. at 2050-51 .

While we do not suggest that height restrictions can never be verified, the evidence introduced in this case was inadequate to prove that the height restrictions were manifestly related to employment by the sheriff’s department. The district court’s finding of job-relatedness must be reversed.

II. § 1981 Claim

Because the case must be remanded for other reasons, we will consider briefly the plaintiffs’ argument that they are entitled to relief under 42 U.S.C. § 1981 .

Plaintiffs claim that the sheriff’s department is liable for employment discrimination under 42 U.S.C. § 1981 , for the challenged employment practices occurring between March 12, 1970 and March 24, 1972, prior to the applicability of Title VII to public employers. Plaintiffs concede that they cannot prove intentional discrimination but argue that employment discrimination claims under § 1981, like Title VII claims, do not require such proof. This is not the law.

The parties recognize that this circuit lacks controlling precedent on the precise issue presented here. However, the historical connection between § 1981 and the Fourteenth Amendment, the thrust of the Supreme Court decision in Washington v. Davis, supra, the complexity as well as legislative history of the statutory scheme adopted by Congress in Title VII, and the ease with which acceptance of plaintiffs’ theory would allow circumvention of that scheme convince us that equal protection rather than Title VII standards apply to employment discrimination claims brought under § 1981. Thus, we affirm that part of the judgment denying relief under § 1981.

Affirmed in part, reversed in part and remanded for further proceedings not inconsistent with this opinion.

. The term “Mexican-American” was defined by the district court as “any person who is ‘Spanish-Sumamed’ or a ‘Spanish Language’ person as those terms are utilized by the United States Department of Commerce, Bureau of the Census, in the 1970 Census.” . Since Title VII did not become applicable to the sheriffs department, a municipal employer, until 1972, plaintiffs complain that selection devices used between 1970 and 1972 were illegal under the Civil Rights Act of 1870. . Approximately 45% of the Mexican-Americans who took the test failed, while 23% of the whites failed. Similar or less dramatic disparities between failure rates have been found to constitute substantial adverse impact. See, e. g., Jones v. New York City Human Resources Administration, 528 F.2d 696, 698 (2nd Cir.), cert. denied, 429 U.S. 825 , 97 S.Ct. 80 , 50 L.Ed.2d 88 (1976), (failure rates ranged from 63% to 85% for Hispanics and 69% to 84% for blacks, compared to 12% to 49% for whites); League of United Latin American Citizens v. City of Santa Ana, 410 F.Supp. 873, 902, 906-07 (C.D.Cal. 1976) (70%, 68.2%, 64.7% failure rates for Mexican-Americans versus 41.9%, 41.4%, 46.3% for whites); Western Addition Community Organization v. Alioto, 340 F.Supp. 1351, 1353 (N.D.Cal. 1972) (80% failure rate for blacks, 43% for whites). . The court’s finding is clearly inconsistent with the stipulated statistics. The district judge appears to have mistakenly lifted the failure rates for oral interviews instead of those for the 5IX test from the same page of the Pretrial Conference Order on which both sets of data appear. . In the instant situation, a perfect positive correlation would mean that every applicant for deputy sheriff who scored high on the ET4-1 examination also performed well in the academic courses at the sheriffs academy and that every candidate who did poorly in the preemployment test later did poorly at the academy. "9 The EEOC guidelines recognize that validation studies should include persons who are excluded by the selection standards in question. (See 29 C.F.R. § 1607.5 )” 595 F.2d at 1380 . . The height requirements in effect until May 1975 (67 inches for males and 63 inches for females) had the effect of eliminating 41% of the Mexican-American males (compared to 14% of all other males) and 74% of the Mexican-American females (compared to 50% of the other females) from further consideration for employment. . Concurring in Dothard , Justice Rehnquist suggested that “the appearance of strength” was a qualification for prison guards justifying minimum height and weight requirements despite their adverse impact on women. 433 U.S. at 339-40 , 97 S.Ct. at 2731-32. . Mere opinion testimony by sheriffs department personnel that height is effective in control functions is inadequate to establish the significant correlation that is required under Title VII. Cf. United States v. Virginia, 454 F.Supp. 1077, 1088 (E.D.Va.1978) (opinion evidence that “height and weight are useful in police work” does not prove job-relatedness, but rather only “a good faith belief of job-relatedness.”) . This court decided the § 1981 issue in favor of the plaintiffs’ position, over a strong dissent, in Davis v. County of Los Angeles, 566 F.2d 1334 (9th Cir. 1977). However, because the Supreme Court subsequently vacated the decision on the ground of mootness, County of Los Angeles v. Davis, 440 U.S. 625 , 99 S.Ct. 1379 , 59 L.Ed.2d 642 (1979), it is no longer binding upon us. . We are not alone in this conclusion. See, e. g-, Mescall v. Burrus, 603 F.2d 1266, 1270-71 (7th Cir. 1979); Williams v. DeKalb County, 582 F.2d 2 (5th Cir. 1978); Johnson v. Alexander, 572 F.2d 1219 (8th Cir.), cert. denied, 439 U.S. 986 , 99 S.Ct. 579 , 58 L.Ed.2d 658 (1978); Chicano Police Officer’s Ass’n v. Stover, 552 F.2d 918 (10th Cir. 1977); Note, Davis v. Los Angeles: Plaintiff’s Burden of Proof Under Section 1981, 9 Golden Gate U.L.Rev. 1, 19-20 (1978-1979); But see Note, Section 1981: Discriminatory Purpose or Disproportionate Impact?, 80 Colum.L.Rev. 137 (1980).