Thus there is counting and counting, distinguishing between low numbers by directly appreciating the difference between two quantities of sensuous perceptions, and distinguishing between numbers of any amount by marking each sensuous perception with a separate sign. Of course, in the above instance of animals counting it must be the former method alone that is employed, and, therefore, I have not sought to carry the ape beyond the number 5 lest I should spoil the results already gained. But a careful research has been made to find how far this method can be carried in the case of man. The experiments consisted in ascertaining the number of objects (such as dots on a piece of paper) which admit of being simultaneously estimated with accuracy. It was found that the number admits of being largely increased by practice, until, with an exposure to view of one second's duration, the estimate admits of being correctly made up to between 20 and 30 objects. (Preyer, 'Sitzungsber. d. Gesell, f. Med. u. Naturwiss.,' 1881.) In the case of the ape it is astonishing over how long a time the estimate endures. Supposing, for instance, that she is requested to find five coloured straws. She perfectly well understands what is wanted, but as coloured straws are rare in the litter, she has to seek about for them, and thus it takes her a long time to complete the number; yet she remembers how many she has successively found and put into her mouth, so that when the number is completed she delivers it at once. After having consigned them to her mouth she never looks at the straws, and therefore her estimate of their number must be formed either by the feeling of her mouth, or by retaining a mental impression of the successive movements of her arm in picking up the straws and placing them in her mouth. Without being able to decide positively in which of these ways she estimates the number, I am inclined to think it is in the latter. But, if so, it is surprising, as already remarked, over how long a time this estimate by muscular sense endures. Should we trust Houzeau's statement, however (and he is generally trustworthy), it appears that computation by muscular sense may extend in some animals over a very long period. For he says that mules used in the tramways at New Orleans have to make five journeys from one end of the route to the other before they are released, and that they make four of these journeys without showing any expectation of being released, but begin to bray towards the end of the fifth.[67]