200 RAYMOND PEARL AND MAUD DEWITT PEARL. 



possible comparison cases is the difference less than four times 

 its probable error. In that case (Spanish-Argentine and Spanish- 

 Spanish) the difference is 1.4 times its probable error, and could 

 not, taken by itself, be considered significant. Taking into 

 account, however, the facts that (a) the difference is of the same 

 sense as the other differences in the table and (/-') that it is larger 

 than its probable error the general conclusion reached from the 

 other figures is not vitiated. This conclusion is that within the 

 limits of the present material t/iar is evidence of the significantly 

 greater proportionate production of males iji the offspring from 

 matings involving different racial stocks than in the offspring from 

 matijigs in zvhich both parents belong to the same racial stock. 



Discussion. 



The data which are set forth in the tables given above appear 

 to lead clearly to the conclusion which has been drawn from 

 them. This conclusion, however, is merely a statement of fact. 

 In interpreting it it remains to consider two points. The first of 

 these is as to whether there are limitations or fallacies in the 

 data themselves which invalidate the conclusion to which they 

 appear to lead. The second is as to what is the meaning of the 

 facts implied by this conclusion supposing it to be true. One 

 cannot be too cautious in drawing conclusions from human vital 

 statistics of whatever kind. Vital statistics notoriously abound 

 in pitfalls. In a critical examination of the data with a view to 

 possible criticism and interpretation the following points suggest 

 themselves : 



I. That the material is not sufficiently extensive. It might 

 conceivably be maintained that if a larger number of births were 

 to be dealt with they would show a different result. For two 

 reasons such a consideration appears to have little weight. In 

 the first place the number of births included in the statistics is 

 extremely large as measured by biological standards. The sta- 

 tistics include upwards of 200,000 births. In the second place 

 the probable errors of the sex-ratios indicate how literally enor- 

 mous are the combined odds against such a consistent system of 

 differences as that shown in Table III., arising fortuitously. In 

 this connection it may be said that the work was begun in the 



