THE FACTOR OF IMMUNIZATION I93 



If we apply this formula to the mean values of Table I (or Fig. i) 

 we will get 



I F I. = W 3 (i-^7 + i-74 + 2-52)-(i-27 + i'97+2-54) 



3x 1-32 



= o-477 + ^-^ = o-540,orF.I. = 3-5. 



This figure, as seen from Table I, is an estimate based on 

 pooling the results from experiments in which the history of the 

 immune donor varies from having been used at 7 days after i 

 injection to 8 days after 5 injections. While this procedure is fully 

 legitimate for calculation of the common slope, b, it is, neverthe- 

 less, wasteful of information. 



We also want to know if there is any variation in the F.I. 

 depending on the schedule of immunization. Tentatively, the F.I. 

 has been calculated for each test litter (inserting the corresponding 

 mean responses in the formula above, and using the '*semi- 

 platonic" slope throughout). These figures are entered in the last 

 column of Table I. Apart from a single test litter (12 17) there is no 

 more than a rather feeble suggestion that the F.I. is dependent on 

 the number of immunizing injections. The high factor given by 

 litter no. 1217 is most hkely to be due to an exceptionally great 

 sampling error, all the more so since parallel titration of the 

 same suspension in litter no. 12 19 gave a F.I. within the normal 

 range. 



The procedure of determining the F.I. for each individual litter 

 is not, however, the best way to treat the data — not even if they 

 were obtained from assays with fully comparable conditions. 

 The sampling error is just too big, as illustrated above by the test 

 litters no. 1217 and 1219. 



It seems more reasonable to subdivide the titrations of Table I 

 into two groups according to whether the immunized donor has 

 received a single or repeated injections, and thereafter treat each 

 group as a whole. The group representing single injections 



