364 PHYSICAL PROPERTIES 



which is the same form as (i) . Obviously, however, in a solution of 

 a protein in a solvent the density or refractivity of which departs 

 at all widely from those of water, although the law n — rii = aX c 

 might be expected to hold good, the constant a would not have 

 the same value as in aqueous solvents. In other words, in non- 

 aqueous solvents the law n — rii = aX c should hold good, but, 

 for a given protein the value of a should vary with the nature of 

 the solvent depending upon its refractivity and its density. In 

 satisfactory correspondence with this deduction we find that the 

 law n — Til = aX c holds good for solutions of potassium casein- 

 ate in alcohol-water mixtures containing from to 75 per cent of 

 alcohol, for solutions of serum-globulin in alcohol-water mixtures, 

 and for solutions of gliadin in a variety of solvents. As we shall 

 see, however, although the value of a for a given protein varies 

 with the nature of the solvent in the sense demanded bj'^ the above 

 theory, this variation cannot be quantitatively expressed by the 



formula 100 a = iV ^ -A^i + 1 - ^• 



a a 



In estimating the accuracy with which a can be determined in 

 the equation n — wi = a X c, it is essential to recollect that the 

 experimental error in determining the angle of total reflection with 

 the Pulfrich Refractometer is ±1'. This, for the instruments 

 which I have employed, corresponds to an error of ±0.00009 in 

 the determination of the refractive index of a given solution. Now 

 it is obvious that since the absolute error in the determination of 

 n — ni is the same, the error in the determination of a must be 

 less in proportion to the magnitude of c. 



In order to assign to each determination its due weight in the 

 estimation of the mean value of a for any solvent we must therefore 

 add together all of the observed values of n — Wi and divide this 

 sum by the sum of the concentrations employed. This procedure 

 has been adopted in my own estimates, cited below. The follow- 

 ing are the values of a for various proteins and solvents which 

 have so far been ascertained. The influence of temperature upon 

 the magnitude of a appears to be very slight. In fact for casein I 

 have found the value of a to be the same, within the experimental 

 error, at 40 degrees as it is at 20 degrees. Hence the temperatures 

 at which the observations were made are not specified in the 

 accompanying table; they lay, in every case, however, between 

 ordinary room-temperatures and 24 degrees. 



