So far, then, we have highly satisfactory evidence of tissue-tracts performing the function of afferent nerves. But another point of interest here arises. Although, in the experiment just described, the manubrium is no longer able to _localize_ the seat of stimulation in the bell, it nevertheless continues able to perceive, so to speak, that stimulation is being applied in the bell _somewhere_; for every time any portion of tissue below the cut a is irritated, the manubrium actively dodges about from one part of the bell to another, applying its extremity now to this place and now to that one, as if seeking in vain for the offending body. If the stimulation is persistent, the manubrium will every now and then pause for a few seconds, as if trying to decide from which direction the stimulation is proceeding, and will then suddenly move over and apply its extremity, perhaps to the point that is opposite to the one which it is endeavouring to find. It will then suddenly leave this point and try another, and then another, and another, and so on, as long as the stimulation is continued. Moreover, it is important to observe that there are _gradations_ between the ability of the manubrium to localize correctly and its inability to localize at all, these gradations being determined by the circumferential distance from the end of the cut and the point of stimulation. For instance, in Fig. 22, suppose a cut A B, quarter of an inch long, to be made pretty close to the margin and concentric with it, then a stimulus applied at the point _c_, just below the middle point of A B, would have the effect of making the manubrium move about to various parts of the bell, without being able in the least degree to localize the seat of irritation. But if the stimulus be applied at _d_, the manubrium will probably be so far able to localize the seat of irritation as to confine its movements, in its search for the offending body, to perhaps the _quadrant_ of the bell in which the stimulation is being applied. If the stimulation be now supplied at _e_, the localization on the part of the manubrium will be still more accurate; and if applied at _f_ (that is, _almost_ beneath one end of the cut A B), the manubrium may succeed in localizing quite correctly.