The above differences are more significant than would at first appear. An error of 10 months in the mental age of a 5-year-old is as serious as an error of 20 months in the case of a 10-year-old. Stating the error in terms of the intelligence quotient makes it more evident. Thus, an error of 10 months in the mental age of a 5-year-old means an error of almost 15 per cent in the intelligence quotient. A scale which tests this much too low would cause the child with a true intelligence quotient of 75 (which ordinarily means feeble-mindedness or border-line intelligence) to test at 90, or only slightly below normal. Three serious consequences came from the too great ease of the original Binet scale at the lower end, and its too great difficulty at the upper end:-- 1. In young subjects the higher grades of mental deficiency were overlooked, because the scale caused such subjects to test only a little below normal. 2. The proportion of feeble-mindedness among adult subjects was greatly overestimated, because subjects who were really of the 12- or 13-year mental level could only earn a mental age of about 11 years. 3. Confusion resulted in efforts to trace the mental growth of either feeble-minded or normal children. For example, by other versions of the Binet scale an average 5-year-old will show an intelligence quotient probably not far from 110 or 115; at 9, an intelligence quotient of about 100; and at 14, an intelligence quotient of about 85 or 90. By such a scale the true border-line case would test approximately as follows:-- At age 5, 90 I Q (apparently not far below normal). At age 9, 75 I Q (border-line). At age 14, 65 I Q (moron deficiency). On the other hand, re-tests of children by the Stanford revision have been found to yield intelligence quotients almost identical with those secured from two to four years earlier by the same tests. Those who graded feeble-minded in the first test graded feeble-minded in the second test: the dull remained dull, the average remained average, the superior remained superior, and always in approximately the same degree.[18] [18] See "Some Problems relating to the Detection of Border-line Cases of Mental Deficiency," by Lewis M. Terman and H. E. Knollin, in _Journal of Psycho-Asthemes_, June, 1916. It is unnecessary to emphasize further the importance of having an intelligence scale which is equally accurate at all points. Absolute perfection in this respect is not claimed for the Stanford revision, but it is believed to be at least free from the more serious errors of other Binet arrangements. CHAPTER V ANALYSIS OF 1000 INTELLIGENCE QUOTIENTS An extended account of the 1000 tests on which the Stanford revision is chiefly based has been presented in a separate monograph. This chapter will include only the briefest summary of some of those results of the investigation which contribute to the intelligent use of the revision.