I asked all the subjects before they made the real experiment to distribute 24 other cards in 4 piles, on each of which one of the four letters, A, E, O, and U was printed only once.  Hence no comparison of various factors was involved in this form of distribution.  The average time for this ordinary sorting was about 20 seconds.  Only rather quick individuals carried it out in less than 18 and only very slow ones needed more than 25 seconds.  This maximum variation of 10 seconds is evidently insignificant, as the variations in the experiment amount to more than 200 seconds.  But it is very characteristic that the results of the two experiments do not move parallel.  Some persons, who are able to sort the cards on which only one of the 4 letters is printed very quickly, are rather slow when they sort the cards with the 48 letters for which the essential factor is the act of comparison.  In the first case the training in card-playing also seems to have a certain influence, but in the second case, our real experiment on decision, this influence does not seem to exist.

We have emphasized from the start that it is no less important to give consideration to the number of mistakes.  A mere rapidity of distribution with many mistakes characterizes, as we saw, a mental system which is just as unfit for practical purposes as one which acts with too great slowness.  But it would not have been sufficient simply to ask how many cards were put into wrong piles.  The special arrangement of the cards with four different types of combinations was introduced for the purpose of discriminating among mistakes of unequal seriousness.  When one letter appeared 21 times and the three others only 9 times, it was surely much easier to make the decision than when the predominant letter appeared only 15 times and the other three each 11 times.  The easier the right decision, the graver the mistake.  Of course the valuation of these mistakes must be rather arbitrary.  We decided to value as 4 every mistake in these cards on which the predominant letter appears 21 times; as 3, a mistake in the 18 letter cards; as 2, a mistake in the 16 letter cards; and as 1, a mistake in the most difficult ones, the 15 letter cards.  If the mistakes are calculated on this basis and are added together, a sum below 5 may indicate a very safe and perfectly reliable ability for decision; 5 to 12, satisfactory; 12 to 20, uncertain; and over 20, very poor.  In order to take account of both factors, time and mistakes, we multiply the sum of the calculated mistakes by the number of seconds.  If the product of these two figures is less than 400, it may be taken as a sign of perfect reliability in making very quick, correct decisions, in complex life situations; 400 to 1000 indicates the limits between which the ability for such decisions may be considered as normal and very satisfactory; 1000 to 2000, not good but still adequate; 2000 to 3000, unreliable, and over 3000, practically absent.  It is clear