20 



BALD 



Table I 



obtained by adding or subtracting to the terms of series 1 randomly 

 chosen percentage variations of the magnitude shown beside series 3 

 and rounding off the terms to the nearest whole numbers. In fitting 

 series 3, the value of A^ was assumed to be 98 instead of 100 and the 

 value of pn at the highest concentration of virus was assumed to be 

 23.04 (series 4), instead of 23.89 as in series 2. The fit obtained would 

 be considered very good if 3 were an experimental series. 



The deviation from a Poisson distribution in series 1 is trivial com- 

 pared with deviations in numbers of local lesions found in dilution 

 trials with many samples of tobacco mosaic and attributed to aggrega- 

 tion of virus particles. No amount of adjustment will give Poisson 

 series in agreement with the numbers of local lesions found in such 

 trials. The hypothesis of virus aggregation was applied to these cases. 

 The equation formulated to cover them is: 



y = NCi — e-P^) (2) 



