The Significance of the Binet-Simon Tests
The Psychological Clinic Copyright, 1910, by Lightner Witmer, Editor. Vol. X, No. 5 October 15, 1916 :Author: Edward L. Thorndike, Ph. D.,
Teachers College, Columbia University, New York In a study of the intellectual status of the dependent children of a certain county, made by Mr. Stenquist, Mr. Trabue, and the author, four series of tests were used?the Binet-Simon series as arranged by Goddard, a series of 56 sentences with omitted words to be filled, a series of passages to be read with questions to measure the comprehension gained by the reading, and a series of mechanisms to be put together out of their loose parts, the mechanisms themselves being at hand for examination.* Each child was first given as a measure in each test the age of the median ordinary child attaining the score which the dependent child in question attained. Then, as his final measures, the four differences (plus or minus) of these ages from the dependent child’s actual chronological age were used. It is of some theoretical interest, and of great practical importance, to measure the correlations between these four measures of superiority or inferiority to the median ordinary child of the same chronological age. I have therefore computed the Pearson coefficients of correlation with the following results:
the interrelations of the three tests of intellect involving ability WITH WORDS, NUMBERS, AND THE LIKE. 29 10-year olds. 31 11-year olds. 35 12-year olds. 35 13-year olds. 23 14-year olds. Average of the 5 r’s. Binet with Completion Test. .86 h .80| .89 .94 .87 .87-2 Binet with Reading Test. .78 .70 .84 .90 h .82 .81 Completion with Reading Test. .74 .82 .90 .92 .92 .86
It is clear that these three series measure much the same complex of abilities and achievements. If the Binet tests measure general intellect, so to a high degree must the tests in merely reading a passage with understanding. The Binet tests are almost as much like intelligent reading in what they measure as the completion tests, dealing exclusively with words, are. If the Binet tests are really adequate to measure that mixture of ability to deal with words and with facts expressed in words and other symbols, ability to deal with things of all sorts and ability to deal with people, whereby a person lives successfully, then the very much more convenient Completion and Reading tests are also approximately accurate measures of that mixture and should be used widely in mental diagnosis. If, on the other hand, as I think, the Binet test measures chiefly the ability to deal with ideas expressed in words, and is inadequate as a measure of mechanical ability or executive ability, the fate of children should not be decided so exclusively by it, as is now the case. This latter hypothesis is supported by the correlations in the case of the test of mechanical ability, or construction test. If the Binet tests measure general ability adequately they should correlate closely with the construction test?as closely, say, as they do with the reading test (except for a possibly greater “attenuation” due to a greater unreliability in the construction test than in the reading test). They do not. On the contrary, they show no superiority in this respect to the completion test and very little superiority to the reading test. The correlations found (Pearson coefficients) are as follows:
RELATIONS OF THE BINET, COMPLETION AND READING SCORES TO THE CONSTRUCTION TEST SCORES. 28 10-year olds. 11-year olds. 12-year olds. 13-year olds. 24 14-year olds. Average of 5 r’s .. Binet with Construction. .68 ? 74| .55| ? 56| .604 .63 Completion with Construction. .68 .70 .62 .511 .731 .65 Reading with Construction. ? 55| .70 .40 h .45| .75h .571 Composite of Binet, Completion and Reading with Construction. .66 .76 .59 ? 55| .75 .66
Clearly the Binet test is not an adequate test of intellect in general for even very stupid children of ages 10-14. It tests much the same ability or complex of abilities as the completion and reading tests. This ability or complex of abilities is far from identical with the ability measured by the construction test, for example. It should be noted that, although it is fair to use the relative amounts of these correlations for the purposes of the argument of this paper, their absolute amounts should not be assumed to hold good for a group formed by random sampling. These dependent children are inferior to average children of equal chronological age in the construction test, and are very much inferior to average children in the other three tests. Their deviations being taken from the average child of the age in question, we have a selection largely of minus deviations, and it is not certain that the correlation calculated from such a selection will represent fairly the correlation calculated from a random sampling of all.
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.