ALCOHOLISM AND BEHAVIOR OF WHITE RATS 



239 



The fact of there being a series of twelve comparisons, eleven 

 of which go in the same direction, has, in itself, a significance 

 independent of the probable errors obtained in each case from 

 the standard deviations. There is presented a series of twelve 

 terms; each term may be plus or minus. If the sign is due to 

 chance, there should be as many plus as minus terms in the 



TABLE 4 

 Showing the significance of the difference between the averages of the time of tests 

 and controls on the different days of their training. These differences are taken 

 from the averages of sexes and strains together in table 3. The plus sign indicates 

 that the average of the test is higher than the corresponding average of the control. 

 Since the differences in the first part of training are greater than three times their 

 probable errors, the tests appear to take a significanthj longer time in the maze in 

 this part of training 



Training. 



Retention . 



'completes' and 'lNCO>rPLETES,' 

 AND 'failures' 



Diff. 



P. E. Diflf./P. E. 



+530.9±130.2 

 +228. 9± 56.7 

 + 110. 5± 44.0 

 + 50. 2± 14.3 

 + 14. 6± 9.9 

 + 9.3± 9.5 

 + 16. 6± 8.3 

 + 7.7± 7.3 



- 5.5± 8.3 



+ 0,2± 6.2 



+ 13. 7± 5.5 



+ 30. 4± 6.9 



4.08 

 4.03 

 2.51 

 3.50 

 1.46 

 0.98 

 2.00 

 1.06 



0.66 

 0.03 

 2.50 

 4.40 



long run. The chances that eleven out of twelve should be plus 

 on a chance basis would be the same as the frequency of eleven 

 heads when twelve coins are tossed simultaneously; departures 

 from equality as great as this may be expected to occur once in 

 157 trials, which corresponds to a little over four times the 

 probable error. 



When each day of training is considered by itself, the averages 

 in the great majority of cases show that the tests took more 



