After securing such standard values, we carry on the experiments in the following form. Six different postal cards, for instance, are seen on a black background through the opening of a shutter which is closed after 5 seconds. The six may be made up of a landscape, a building, a head, a genre scene, and so forth. After 20 seconds the same group of postal cards is shown once more, except that one is replaced by a similar one, instead of one church another church building, or instead of a vase with roses a vase with pinks. If the substituted picture has the average similarity value of 80 per cent and we make the experiment with 10 persons, the substitution may be discovered by 7 persons and remain unnoticed by 3. We can now easily vary every one of the factors involved. If instead of 6 cards, we take 10, it may be that only 4 out of 10 persons, instead of 7, will discover the substitution, while if we take 4 cards instead of 6, perhaps 9 persons out of 10 will recognize the difference under these otherwise equal conditions. Only an especially careless observer will overlook it. But instead of changing the number of objects, we may change the periods of exposure. If we show the 6 cards only for 2 seconds instead of 5 seconds, the number of those who recognize the difference may sink from 7 to 5 or 4, and if we make the time considerably longer, we shall of course reach a point where all 10 will recognize the substitution. The same holds true of the shortening or lengthening of the time-interval between the two presentations. The third variable factor is the similarity itself. If instead of one church, not another church, but a theatre or a skyscraper is shown, that is, if the similarity value of 80 per cent sinks down to a similarity of 60 per cent or 50 per cent, the number of those who recognize the substitution will again become larger; if, on the other hand, the substituted card shows the same church, only from a slightly different angle, bringing the similarity value up to 90 per cent or 95 per cent, the number of observers who recognize the substitution may sink to 2 or 3. To make the experiments reliable, it is also necessary frequently to mix in cases in which no substitution at all is introduced.