The Status of Tests for the Measurement of Clerical Aptitude
- Author:
Roy N. Anderson
Associate in Guidance and Personnel, Teachers College, Columbia University The past two decades have brought forth considerable activity in the measurement or testing of a variety of supposed personal traits and, from the results of these tests, predictions are being made regarding the possible achievements of the individuals tested. Among the proposed tests a considerable number have been devised for the detection of “clerical aptitude”. In combing the literature over sixty articles on such tests have been found. The length and content of the articles range from a few pages of general information with a title “Empty Heads Make Poor Clerks” to a volume entitled “A Diagnostic Test of Aptitude for Clerical Office Work.” Of the articles in this group only fourteen present data relative to the specific testing procedures involved in the measurement of clerical aptitude.2
Just because a test is called a test of clerical aptitude is no guarantee that it measures such aptitude. No matter how interesting the title of the test sounds, we must critically examine its fundamental concepts and structure. We must insist that it conform to the requirements that embrace scientific procedure. Fortunately these scientific principles are rather clearly defined. Let us enumerate them and see how this group of so-called tests of” clerical aptitude” meet their requirements. It is generally agreed that before a test can be shown to measure a given aptitude, it must be given to persons known to possess that aptitude, and their standings in the test must correspond with their working proficiency. The question that always perplexes the investigator is: What shall I use an an index of their proficiency? In technical terms, what shall be the criterion?
The criterion, according to Burtt, “is an index of occupational proficiency which is used in evaluating the tests designed to predict that proficiency.”3 If the criterion is inadequate then the test can 1 An address given at the meeting of the National Vocational Guidance Association, Feb. 21, 1930, Atlantic City, N.J.
2 For full description of tests see Anderson, Roy N. Measurement of Clerical Ability?The Personnel Journal, Vol. VIII, No. 4, Dec. 1929. 3 Burtt, Harold E., Employment Psychology, Houghton Mifflin Company, New York, 1926, p. 169. have no proven validity. A valid criterion is admittedly hard to obtain. What should one use as a measure of the proficiency of a clerical worker? When we attempt to define clerical work we are amazed at its ramifications, ranging from the activities performed by the humble file clerk to the duties performed by the expert secretary. One investigator found thirty different titles assigned to clerks. With such a diversity of activities and unstandardized procedures it is extremely difficult to secure measures of proficiency. Hull lists three types of aptitude criteria4: 1. Product; for example, the number of entries made in a ledger. 2. Action; for example, the time required to file a given number of cards. 3. Subjective impression. This is generally obtained by having the workers rated by one or more superiors.
Let us examine the so-called tests of “clerical aptitude” and see whether they fulfil the demands of scientific procedure. Though there were in the literature examined fourteen tests, four of this number did not publish information concerning their criteria of vocational success. Of the remaining ten, one test utilized a criterion consisting of production: efficiency was determined in terms of speed and accuracy. Eight tests used criteria which Hull would call subjective impressions, the ratings being made in most cases by department heads or supervisors. One test utilized scores on other performance tests.
The second respect in which we should examine these tests is in regard to their reliability. That is, does an experimenter obtain the same results every time he administers the test? Since a test is at best only a sample of the aptitude which is being measured, this sample should be as representative as possible. The measure of reliability is generally given in terms of a coefficient of correlation. The correlation coefficient should be .90 or above. Only two of the fourteen investigators whose tests were examined reported an index of reliability. In one case it was .82 + .02 and in the other investigation reliabilities of .49 (one test) and .26 (in another) were reported. It is inconceivable that this fundamental procedure was not followed out in the construction of the other twelve tests. The third fundamental requirement in scientific method is to ascertain the validity of the test. This is determined by comparing the scores received on the test with the scores received on the criterion. One must prove that the persons who stand high in the test 4 Hull, Clark L., Aptitude Testing, World Book Company, Yonkers, New York, 1928, pp. 375-376.
stand high on the criterion, which is supposed to be a measure of occupational proficiency, and that those who stand low in the test stand low on the criterion. The method usually employed is to compute some index of correlation. It is generally expressed by the Pearson r, but this measure is valid only if the relationship is rectilinear. With a curved regression line the correlation ratio or eta is the preferable method. There is a misunderstanding as to how high the correlation coefficient should be in order to be “significant.” Some investigators point to a correlation coefficient of .50 as “high” where as it may be really only “significant.” Statisticians generally require a correlation of .70 or more to indicate a high degree of relationship. Table I shows the size of the various correlations reported:
Table I 5 tests had correlations between 30-40 6 tests had correlations between 40-50 1 test had correlations between 50-60 5 tests had correlations between 60-70 2 tests had correlations between 70-80 1 test had correlations between 80-90 Here we see that only three of the correlation coefficients are above the minimum generally required.
The validity of a test is dependent not only on the size of the correlation coefficient, but also upon the reliability of the coefficient. This in turn depends on the number of cases used. Some of the investigators do not even report the number of cases on which they standardized the test. Of those listed, the number of cases are 43, 50, 55, 90, 100, 188, 200. The number in some of the investigations is probably so small as not to constitute an adequate sampling. Statisticians say that an investigator should furnish figures regarding the “probable error” of his data in order to permit others to check and verify the findings, and also to enable other investigators to determine what the chances are that any other sample taken at random will fall within the proper limits. In order that we may be sure there is some correlation present, the coefficient of correlation should be at least four times its probable error. Only four of the studies listed “probable errors” with their findings. In discussing the use of correlations in examining the validity of a test, we might call attention to a misunderstanding many people seem to have regarding a high correlation. They assume that it is indicative of causal relationship supposing, that because two things show a high correlation, one is the cause of the other. Such a dependence does not necessarily hold true. Granting that there is dependence, there is always the possibility that there is a third factor on which both depend and hence even with a high correlation, we dare not assert that a test really measures the aptitude for which we are searching.
Summary
After this careful scrutiny of the clerical tests which have been recommended for use, one is impressed with the valiant attempts that have been made, but one is obliged to admit that most of them have failed to produce tests which are scientifically defensible. The vocational counselor who desires to use these tests will find them of little or no value. We should, however, distinguish between tests for the measurements of “aptitude for clerical work and tests for the measurement of proficiency in clerical work. “An aptitude test is a test designed to discover what potentially a given person has for learning some particular vocation or acquiring some particular skill.” (Hull)5 AVhile not one of these tests can lay clear claim to be an aptitude test, some of them, which may have been designed to measure acquired skill, may perhaps be legitimately classed as proficiency tests.
A further distinction should be drawn between the value of tests in vocational guidance?helping a person to choose a vocation, and vocational selection?selecting a worker for a job. A few of these tests may be useful in vocational selection (probably because they measure learned skills), but that does not render them equally valuable in vocational guidance and it does not prove that they detect “clerical aptitude.” 6 Clark L. Hull, op. cit., p. 50.
Disclaimer
The historical material in this project falls into one of three categories for clearances and permissions:
Material currently under copyright, made available with a Creative Commons license chosen by the publisher.
Material that is in the public domain
Material identified by the Welcome Trust as an Orphan Work, made available with a Creative Commons Attribution-NonCommercial 4.0 International License.
While we are in the process of adding metadata to the articles, please check the article at its original source for specific copyrights.