ALCOHOLISM AND BEHAVIOR OF WHITE RATS 231 



number below each day is the number of times the difference 

 between the test and control averages is greater than its probable 

 error, considering just the 'completes' and 'incompletes.' The 

 data for this figure are given in table 3. 



The first half of table 3 is based on the 'completes' and 'incom- 

 pletes,' the second half, at the right, includes the 'failures' as 

 well. 'Completes' averaged alone are not given; they show 

 substantially the same results as when averaged with 'incom- 

 pletes.' In each half of the table there are three main sections: 

 1, males; 2, females; 3, males and females (i.e., primary data 

 put together regardless of sex, not averages of the averages of 

 each sex). Under each of these headings the time per rat for 

 the tests and controls is given for each day (eight days of training 

 and then four days of retention). The three strains are con- 

 sidered separately, then all strains together. The number of 

 rats in each series of averages is shown at the top of the series. 

 As before, the averages are compared by ratios; plus ratios 

 indicate that the tests took longer time than the controls; minus 

 ratios indicate that the controls took longer time than the tests. 



If just the signs of the ratios are studied, it is obvious that 

 there is a great preponderance of 'plus' signs. When the 'com- 

 pletes' and 'incompletes' are considered, the males give 44 'plus' 

 ratios out of 48; the females, 38 'plus' ratios out of 48; and the 

 males and females together, 40 'plus' ratios out of 48. When 

 the 'failures' are included, out of 48 cases the males give 42 

 'plus' ratios, the females 37 'plus' ratios, and the males and 

 females together 40 'plus' ratios. In the whole table there are 

 241 'plus' ratios out of a possible 288. This is, of course, only 

 a general comparison, since in many of these averages the same 

 data are repeated, but it indicates that combining the data in 

 different ways gives the same result. In other words, the greater 

 time taken by the tests is a real difference and not due to the 

 special method of treating the data. Some of the ratios of 

 tests vs. controls given in table 3 are shown graphically in figure 

 4; the sexes together, each strain separately, and all strains 

 together. The straight horizontal line represents equality 

 between the test and control average, or the ratio of 1.000. 



