LIMITS OF HEREDITARY CONTROL 907 
demarkation ‘of embryonic primordia is concerned; but the absence 
of a visible line of separation between the quadrants of the vesicle 
does not preclude the possibility that each quadrant may be 
formed exclusively, or nearly so, of cells derived from one of the 
first four blastomeres. It seems very likely, in fact, that the 
cleavage products of each blastomere would continue to occupy 
the relative position held originally by the parent cell, in spite 
of the various complex developmental processes that ensue, just 
as the cells derived from the first two blastomeres, in many organ- 
isms, retain their bilateral positions and go to form the right and 
left sides of the individual. To this extent then we are probably 
justified in tracing back each of the quadruplets to one of the first 
four blastomeres. If this be granted it results logically that each 
pair must be the product of one of the first two blastomeres. 
On this basis alone are we able to offer any reasonable explana- 
tion of the observed phenomenon of pairing or to attempt an 
answer to the question: Why should there be a closer resemblance 
between paired than non-paired individuals? Our original 
explanation of the condition naively assumed that the first cleav- 
age would divide the fertilized egg into two somewhat unequal 
parts, and that the second cleavage would divide these half eggs 
into quarters more nearly equal than were the halves. In brief 
we assumed that the first cleavage gave products more variable 
than the second. Is there any basis for such an assumption? 
Is there, as development proceeds, a progressive decrease in the 
variability, real or potential, of daughter cells? It has been dis- 
covered from a study of the intra-individual variability of certain 
plants, notably Ceratophyllum, (Pearl, ’10), in which whorls 
of leaves are successively produced by-the apical bud or growing 
point, that the leaves of the first whorl are the most variable, 
least closely correlated, and that later ones vary less in an orderly 
progression. Similarly, if we consider that in the first cleavage 
the armadillo ovum divides itself into two potential individuals 
and that each of these in turn divides into two more individuals, 
we would expect to find a decreasing variability among the like 
parts produced at each successive division. In this case, however, 
the production of quadruplets destroys the twins and we can dis- 
