We have here an instance of erroneous testimony by a witness to his own actions; testimony to the actions of other people is usually much less trustworthy. In this matter of direct perception we may discriminate still further. Tactile perceptions are of almost no value whatever. If a witness be blindfolded and asked to determine, by touch alone, the nature, the volume and the material of an object, it will be found that the responses are very inaccurate. Experience of tactile sensations is relatively small and deductions therefrom are practically valueless. Thus the shock of a bullet entering the body may be interpreted as a slight blow, several dagger thrusts in the back as one thrust, and so on. A piece of ice drawn across the neck of a blindfolded man and warm water simultaneously poured on his chest have been stated to cause death by fright, the man having previously been informed that he was going to have his throat cut. Perceptions of odour or of taste are even less trustworthy; and here the difficulty of expression in precise terms, in the lack of a precise vocabulary, is complicated by the fact that the witness primarily perceives odours and tastes as pleasant or unpleasant, and pays attention only to that aspect of them. In cases of poisoning, therefore, evidence of this kind should be given very little value. It is only when we come to the senses of hearing and of sight that we enter the region where perceptions may have evidential value. In the case of hearing, however, we must still proceed very warily. Experiment has shown that estimates of direction, for example, are quite valueless, since the different estimates made by different observers obey the laws of pure chance. Training can do a little, but very little, to render these perceptions more trustworthy; in general, however, evidence as to the direction of a sound may be neglected. Estimates of the distance of a sound, also, are of very small value. The intensity of a heard sound depends on the intensity of the source, and also on its distance; and these two factors may be apportioned by the observer in the most arbitrary manner. In the case where the sound is articulate, as in overhearing a conversation, we are in the presence of still other sources of error, due to illegitimate inference and the association of ideas. For words which are not heard will be supplied by the witness in all good faith. He will have a theory of the purport of the conversation, and will arrange the sounds he heard to fit it. Edgar Allen Poe’s example, in _The Murders in the Rue Morgue_, of the cries of an ape which were interpreted as remarks in different languages by different observers, is judged by Dr. Locard to be not at all fantastic. The same general source of error applies to visual perceptions. Not everything is observed, and the lacunæ are filled in by the witness in what seems to him the most probable manner. Oversights in proof-reading furnish a familiar example of this kind of error.