Guiding Principles In Choice And Arrangement Of Tests
In choosing his tests Binet was guided by the conception of intelligence which we have
set forth above. Tests were devised which would presumably bring
into play the various mental processes thought to be concerned in
intelligence, and then these tests were tried out on normal children of
different ages. If the percentage of passes for a given test increased
but little or not at all in going from younger to older children this
test was discarded. On the other hand, if the proportion of passes
increased rapidly with age, and if children of a given age, who on other
grounds were known to be bright, passed more frequently than children of
the same age who were known to be dull, then the test was judged a
satisfactory test of intelligence. As we have shown elsewhere,[13]
practically all of Binet's tests fulfill these requirements reasonably
well, a fact which bears eloquent testimony to the keen psychological
insight of their author.
In arranging the tests into a system Binet's guiding principle was to
find an arrangement of the tests which would cause an average child of
any given age to test "at age"; that is, the average 5-year-old must
show a mental age of 5 years, the average 8-year-old a mental age of
8 years, etc. In order to secure this result Binet found that his data
seemed to require the location of an individual test in that year where
it was passed by about two thirds to three fourths of unselected
children.
It was in the assembling of the tests that the most serious faults of
the scale had their origin. Further investigation has shown that a great
many of the tests were misplaced as much as one year, and several of
them two years. On the whole, the scale as Binet left it was decidedly
too easy in the lower ranges, and too difficult in the upper. As a
result, the average child of 5 years was caused to test at not far from
6 years, the average child of 12 years not far from 11. In the Stanford
revision an effort has been made to correct this fault, along with
certain other generally recognized imperfections.