272 E. C. MACDOWELL AND E. M. VICARI 



numbers of errors made. Six types of errors have been recog- 

 nized: type 1, passing a door that should be entered; type 2, 

 turning the wrong way upon entering a door; type 3a, turning 

 back on the correct path; type 3b, turning back on the wrong 

 path; type 4a, retracing the correct path through a door; type 

 4b, retracing the wrong path through a door. About 70 per 

 cent of the errors made are of types 1 and 2. No clear differ- 

 ence between tests and controls was found on the basis of the 

 percentage occurrence of each type of error in the total number 

 of errors. When the strains and sexes are put together, for 

 eighteen of the twenty-four trials in training, the averages of 

 the tests for errors of type 1 are higher than the corresponding 

 control averages; on twenty-one of the twenty-four trials the 

 averages of the tests for errors of type 2 are higher than the corre- 

 sponding averages of the controls. (These statements are made 

 from curves based on the 'completes.') For errors of type 3a 

 the tests have higher averages on ten trials; for errors of type 3b, 

 the tests have higher averages on eleven trials; for errors of type 

 4a, the tests have higher averages on ten trials; for errors of type 

 4b, the tests have higher averages on seven trials (controls higher 

 on one trial). The last three types of errors occur so infre- 

 quently that in most cases after the first seven trials there is no 

 difference between the averages of the tests and controls. In 

 the initial trial for each type the tests have more errors. If 

 the comparisons in figure 14 had been based upon errors of types 

 1 and 2 alone, the differences would have been accentuated. 

 The conclusion may be drawn that, although there does not 

 seem to be much difference between tests and controls in the 

 number of the more unusual errors, the tests are clearly less 

 successful in eliminating the more persistent errors (types 1 

 and 2). This conclusion is strengthened by the following 

 analysis. 



c. The number of trials before each type of error was eliminated. 

 For the purposes of this comparison, an error was said to be 

 eliminated when it occurred only once in four successive trials; 

 these four trials are included in the number of trials before the 

 error was eliminated. 



