DISCRIMINATIVE RESPONSES TO VISUAL STIMULI 485 



A similar concept, formulated in accordance with the "all or 

 nothing" principle might be substituted for the above. Under 

 favorable conditions the impulse conveyed to the effector over 

 the first system of neural arcs excited might be adequate to 

 arouse reaction. Under less favorable conditions the " intensity" 

 of the first impulse alone might be inadequate, but might have 

 to await reinforcement by succeeding impulses conveyed over 

 other pathways which were excited later. Or, a part of the 

 delay may be assumed to have taken place in the central synapses, 

 rather than in the muscle, as was suggested above (p. 464). 



A study of the distribution-curves of the reaction-tunes, not 

 only of one subject, but of several subjects compared with each 

 other, suggests strongly that there are several definite time- 

 values, more or less widely separated, about which the reactions 

 tend to group rather closely. Under the more favorable con- 

 ditions, the preponderance is among the smaller values. 



The foregoing discussion should make it clear that the practice 

 of smoothing curves of distribution of reaction-times by averag- 

 ing the ordinates in overlapping groups of threes is illegitimate, 

 and may also serve to obscure some very instructive features. 



DISPERSION 



The averages of reaction-times obtained under different con- 

 ditions are much more significant when the dispersion of the 

 reaction-times is considered. This measured by the "standard 

 deviation," or quadratic mean of the deviations from the mean. 

 In table 7 is given the standard deviation <r, for each subject 

 under each lighting-condition, with the probable error of the 

 standard deviation; also the differences between the standard 

 deviations of the compared distributions and the probable errors 

 of the differences. The probable error of the standard deviation 

 is derived from the formula 



P*.- 0.6745^ 



and the probable error of the difference between the standard 

 deviations of the compared distributions, from the formula 



