648 R. M. STRONG 
TABLE 8 
Violet sachet powder series 
No: 62 A).0; Bi 2:'C, 1: D; 1 
No. 30 A, 3; B, 0; CG, 0; D, 2 
No. 92 A, 6; B, 0; C, 0; D, 1 
No. 24 A, 2; B, 2; C, 0; D, 0 
~~ 
TABLE 9 
Oil of bergamot series 
INio: 62), AS 352B fase. b45D 24 
NON SO AC O9 a de Oras) 2, 
Now 92) AN 94 Basle Cs 12D 39 
No. 24 A, 31; B, 42; C, 24; D, 44 
In those experiments where A was not the first chamber en- 
tered, the percentage of correct first entrances made by No. 30 
in the oil of bergamot series was 41, and 44 per cent for dove No. 
92. 
It will be seen that doves Nos. 30 and 92 both entered chamber 
A first a very large number of times. This habit became so con- 
firmed in No. 30 that experiments with this bird were finally dis- 
continued. Attempts were made to break up the habit, but no 
suecess was obtained, except when the food was placed regularly 
at one of the three other chambers. This could not be done much, 
of course, without seriously affecting the results of the experi- 
ments. It was desirable that there should be little difference in 
the number of times each chamber was used for food. It will be 
noticed that dove No. 62 did not develop such a habit and that 
No. 24 did not exhibit the tendency in the oil of bergamot series. 
In order to test the possible odor discrimination of the doves 
after they had made one mistake, the means of errors made by 
the birds were calculated. If, for instance, the birds entered 
the chambers at random and did not go into any single chamber 
more than once, they would, in a sufficiently large series of exper- 
iments, be expected to have a mean of errors approximating 1.5. 
As a matter of fact, they often entered an empty compartment 
more than once before finding the food. Thus, on one occa- 
sion, No. 92 entered chambers B and C each three times before 
