3. _Uniformity in the amount and kind of contact with foreign surface._--The capacity of the cannulas was fairly uniform, as stated above; the amount received in them was fairly constant; and the wire hanging in the blood presented approximately the same surface in different observations. A further condition for insuring consistent treatment of the blood in different cases was that of making the tests for coagulation always at the same intervals. Below the writing point of the lever was set an electromagnetic signal _E_, which recorded half-minutes. At the moment a record was made by the signal (see first signal mark, Fig. 25) the clip on the artery was opened, the blood taken, and the process thus begun. In about 20 seconds the cannula was suspended in the water bath and the wire was hanging on the lever. At the next record by the signal and at every subsequent record the vertical rod _R²_ was pushed with the index finger from post _P¹_ to post _P²_ and allowed to move back. This motion was uniform and lasted about one second. The check _R¹_ on the long arm of the lever was thus raised, and as the wire sank in the blood the writing point rose, recording that coagulation had not taken place (see Fig. 25). [Illustration: Figure 25.--Record (reduced two-fifths) of five successive tests of coagulation, with the animal in a uniform condition. The lower line records intervals of 30 seconds. The marks below the time record indicate the moments when the blood samples were drawn.] 4. _Definite end point._--As soon as the blood clotted, the weight of 30 milligrams was supported, and the failure of the lever to rise to the former height in the regular time allowed, recorded that the change had occurred. Very rarely the swing of the lever would be checked for a moment and would then begin to move rapidly, indicating that a strand of fibrin had formed but not sufficiently strong to support the weight, and that when the strand broke, the weight quickly sank in the blood. If this occurred, the next record almost always was the short line, which signified that the weight was well supported. A very slight strand of fibrin was able to prevent the weight from dropping, though at different times the amount of support differed, as shown by the varying length of the final lines (compare first and last series, Fig. 25). These variations are probably a rough indication of the degree of coagulation. In our experiments, however, the length of the final line was disregarded, and merely the fact that the lever failed to swing through its usual distance was taken as evidence of a clot, and the consequent short record was taken as the end point. As soon as this end point was registered, the tube, wire and cannula were lifted out of the bath; the cannula was then separated from the tube and pulled away from the wire. The clot was thus disclosed, confirming the graphic record.