Some of the tests (definitions, comparisons, suggestion) bring out a tendency to automatism which is present in many children. Thus, a child having replied correctly that a butterfly is bigger than a fly, may go on to state that wood is bigger than glass, and paper than cardboard; or having found that "It is a fork" is well received as a definition of that implement, he may give similar replies to the other queries in the definition test. This automatism should not be checked: it should be recorded. The more intelligent children begin to exhibit a certain dissatisfaction with their own answers, however readily they are accepted. It is not always easy to follow the working of the childish mind, and it is not usually advisable to press for further explanation. Such a course is apt to puzzle the child, and render the conditions less favourable. If you are not certain that he should be allowed to pass, you may be practically certain that he should not. Sometimes one gets interesting glimpses into the subject's mentality. A little boy once told me he had never seen a butterfly. Nevertheless, I asked the comparison question, and he gave what is a very usual answer: "A butterfly is bigger than a fly." "How do you know," I said, "if you have never seen a butterfly?" "It's a bigger word," he replied. Another time a little girl, who also declared she had never seen a butterfly, gave another answer which is also very common: "A butterfly is yellow, and a fly is black." The source of this knowledge was not discovered; but one of my students told me later that a child whom she questioned about a butterfly said: "I have seen one; it was blue, but it _ought_ to have been yellow." On being asked why, she responded: "Butter is yellow." The test, of course, is not for the knowledge of the things, but for the power of making a comparison. Occasionally one has to mark a child as doubtful. Thus, in defining abstract terms (XII. 4) Binet records that out of forty-five nine-year-old children, four passed, thirty-six failed, and five were doubtful. This test, however, gives an unusually large percentage of doubtfuls.