Rose Forms as determined by their Cytological Behaviour, 1 8 1 
as a corollary the existence of sexual hexaploid and octoploid roses, if not 
now, then in some past epoch favouring hybridity. 
To postulate sexual forms with such high chromosome numbers 
would thus seem almost unavoidable were it. not that two lines of escape 
remain open, one in the possible occurrence of repeated back crosses, and the 
other in mitotic curiosities like the ‘ mass ’ homotype (PI. X, Fig. 40) observed 
in Rosa Sabini . 
In this fundamental hybridity itself lies one source of the variability of 
the Rosae, since in an ordinary F 1 generation of species hybrids quite 
a wide spread of variation between the conditions of the parents is possible. 
But in the Rosae the spread in any given form (species ?) is far beyond that, 
and hints at that seen in F 2 generations. We have thus to determine 
whether any mechanism revealed in the above investigations allows for 
an approximation to the circumstances of such an F 2 lot. 
In all the descriptions we have given we have emphasized the con- 
stancy of the appearance of seven pairs of chromosomes in the diakinesis, 
and of a curious duplication of the subsequent anaphase ; the anaphase 
involving this paired lot occurred long before that in which the remainder 
took part. So long was the latter delayed that, in many cases, the chromo- 
somes never reached the poles, but either formed micronuclei or were lost in 
the cytoplasm to degenerate there. On the other hand, the associated 
seven uniformly massed themselves as a cap at the poles, and accompanied 
by some of the split univalents formed one of the independent major nuclei 
visible in the interkinesis. At the homotype division these circumstances are 
repeated and exaggerated, with the result that when functional pollen 
is formed, in the great majority of cases, the nuclei of the active grains con- 
tain only seven chromosomes, and these seven originating in the seven 
bivalents of the heterotype division. This was beautifully shown in the 
large number of counts made during a close study of the fate of these 
chromosomes in Rosa coriifolia (type of Fries). 
Since it is merely a matter of chance which member of the original 
homologous fourteen lies on any given side of the equatorial plate, it 
appears from this that we have the same mechanism set before us for the 
segregation of the multitudinous factors of Rosa as is used in explaining 
the segregation of single Mendelizing factors. With seven chromosomes 
taking part, the number of possible combinations is enormous, so that 
gametes carrying them must produce in derived zygotes a huge range of 
genotypes which we see expressed phenotypically in the excessive variation 
of the roses. 
In order that the constancy in the somatic chromosome numbers may 
be maintained the egg-cell in the Villosae must be endowed with twenty- 
one chromosomes and in the rest of the Caninae with twenty-eight, exactly 
as Tackholm has found and as we deduced theoretically. In other words, 
