Most of the actual inferences which we draw during any day belong, indeed, to a much humbler type of thought than do some of the higher forms of non-inferential association. Many of our inferences, like the quasi-instinctive impulses which they accompany and modify, take place when we are making no conscious effort at all. In such a purely instinctive action as leaping backwards from a falling stone, the impulse to leap and the inference that there is danger, are simply two names for a single automatic and unconscious process. We can speak of instinctive inference as well as of instinctive impulse; we draw, for instance, by an instinctive mental process, inferences as to the distance and solidity of objects from the movements of our eye-muscles in focussing, and from the difference between the images on our two retinas. We are unaware of the method by which we arrive at these inferences, and even when we know that the double photograph in the stereoscope is flat, or that the conjurer has placed two converging sheets of looking-glass beneath his table, we can only say that the photograph 'looks' solid, or that we 'seem' to see right under the table. The whole process of inference, rational or non-rational, is indeed built up from the primary fact that one mental state may call up another, either because the two have been associated together in the history of the individual, or because a connection between the two has proved useful in the history of the race. If a man and his dog stroll together down the street they turn to the right hand or the left, hesitate or hurry in crossing the road, recognise and act upon the bicycle bell and the cabman's shout, by using the same process of inference to guide the same group of impulses. Their inferences are for the most part effortless, though sometimes they will both be seen to pause until they have settled some point by wordless deliberation. It is only when a decision has to be taken affecting the more distant purposes of his life that the man enters on a region of definitely rational thought where the dog cannot follow him, in which he uses words, and is more or less conscious of his own logical methods. But the weakness of inference by automatic association as an instrument of thought consists in the fact that either of a pair of associated ideas may call up the other without reference to their logical connection. The effect calls up the cause as freely as the cause calls up the effect. A patient under a hypnotic trance is wonderfully rapid and fertile in drawing inferences, but he hunts the scent backward as easily as he does forward. Put a dagger in his hand and he believes that he has committed a murder. The sight of an empty plate convinces him that he has had dinner. If left to himself he will probably go through routine actions well enough. But any one who understands his condition can make him act absurdly.