1907.] On Globulins. 423 



plotted to scale, the data being derived from Mellanby's paper. They cover 

 only a very small fraction of the whole curve, ABC, but they are sufficient to 

 show, (1) that the curve encloses an area which is placed on the line WS, as 

 in the preceding case. This follows from the form of the curve, and from the 

 fact that if sufficient salt be added, a point is reached on the line WS where 

 .salt solution is in equilibrium with solid globulin. (2) That the area 

 •enclosed increases with a rise of temperature. 



This last fact, together with some observations of my own, enable us to 

 .approximate roughly to an isotherm for edestin at, say, 30°. When edestin 

 is present in sufficient quantity, on warming a further amount passes into 

 solution, so that we have a series of solutions in equilibrium with crystals. 

 Beyond a certain temperature the crystals fuse, and we then have two 

 fluid layers in equilibrium, the lower layer (as I have found by actual 

 -analysis) containing all three components. 



Fig. 3 is a diagram of an isotherm at a higher temperature ; it shows the 

 .area ABC enlarged, and it shows a new area, EOP, due to the appearance of 



••a second series of solutions (the melted crystals). Of the shape of this area 

 •we know nothing. The curve ABC now contains three parts, AD in 

 ♦equilibrium with S, DE with G-, and EF with the solutions which lie on OK. 



Eig. 1 does not represent the conditions for blood globulin at any tem- 

 perature, since, in presence of water alone, a solid solution of water in 

 globulin is formed. The tie lines from the curve BC, therefore, instead 

 of focussing at G, end on a curve of unknown slope, which starts from about 



