ALCOHOLISM AND BEHAVIOR OF WHITE RATS 263 



the group of 'completes.' For the test rats the correlation 

 coefficient of +0.9537 ± 0.0028 was found; for the controls, 

 +0.9607 ± 0.0023, a difference of 0.0070 ± 0.0036 which is 1.9 

 times its probable error. This indicates that the test and con- 

 trol rats agree in showing that the two criteria of time and dis- 

 tance are very closely related; if a rat makes a very poor time 

 record, it generally has a poor distance record, and so if it makes 

 a good time record the distance record is generally good. But, 

 although the correlation coefficients take both these good and 

 poor records into account, there is a difference in the relation 

 in the two cases. It would be impossible for a very short time 

 record to involve a long distance, but a very long time record 

 may give a relatively short distance record; the rat may go very 

 slowly along the correct path, or it may lie down. This means 

 that in one half of the correlation table points may be located 

 almost any^vhere, but in the other half they are greatly restricted. 

 In table 13 is given the part of the combined correlation tables 

 for tests and controls that excludes trials taking more than five 

 minutes. In the terms of the correlation coefficients the time 

 and distance criteria are so nearly alike that they could be con- 

 sidered the same. This similarity is not great enough, however, 

 to remove all the differences in the details of the summaries 

 based on the two criteria separately, but as far as the com- 

 parison of the tests and controls is concerned in the present 

 problem, the two criteria give the same general conclusions. 



We have presented this mathematical proof as well as the 

 empirical demonstration of the similarity of the results given 

 by the two criteria not because there seemed any doubt as to 

 the validity of either method, but to make it perfectly clear 

 to all that the results are entirely independent of the method of 

 treatment. For this same reason we have made the comparisons 

 between the same groups of rats upon still another set of data 

 derived equally from time and distance. 



