26 K. S. LASHLEY 



For these subgroups the correlations are less than for the 

 group as a whole,* as usually follows in any case when the 

 range of one variable is decreased, but they are in all cases 

 positive and fairly large. The analysis, therefore, does not 

 bear out the postulate of a critical mass necessary for the 

 performance of the habit, but rather indicates that deteriora- 

 tion from brain lesion is a continuous function of the extent 

 of the injury. 



4. Munk ('81), Luciani and Seppilli ('86), and other stu- 

 dents of the visual area have held that there is a focal point 

 for visual function in the occipital regien, with surrounding 

 areas of lesser importance, although still concerned in vision. 

 It is possible that in this series of operations some lesions 

 involved such a focal point, others missed it ; that the larger 

 the lesion, the better the chance of including the focal point, 

 and that consequently a greater proportion of the larger le- 

 sions than of the smaller produced loss of the habit and so 

 led to a spurious correlation. To test this, the loci of injury 

 have been subjected to the following analysis. 



Records of the animals which showed the greatest amount 

 of deterioration, making thirty or more errors (75 per cent 

 as many errors as made by them in initial learning) were 

 selected and a composite diagram constructed to determine 

 whether all involved the destruction of a common focal point. 

 These were cases 75, 82, 87, 88, 91, 96, and 97. The composite 

 diagram obtained is given in text figure 4. In it the parts 

 which escaped destruction in some cases but were destroyed 



* By the formula, 



r'xy = l — (1 — rxy)^ 



(Otis, A. S., Jour. Ed. Psychol., 1922, vol. 13, p. 293) the expected correlationa 

 from these data, due to reduction in range only, are : 



For the lower third, r = 0.04 



For the middle third, r = 0.65 



For the higher third, r = 0.58 

 This indicates that the reduction in correlation actually found on splitting up 

 the data is due solely to the reduction in range and that the correlation for 

 the whole group (r = 0.721) represents a continuous function of the group 

 rather than an artifact arising from the inclusion of heterogeneous samples. 



