Harris, Observations on the Physiology of Sced Dcvclopujcnt in Slapltylea. 5 



we expect Vnf to be positive in most cases, and find one fourth 

 only to be negative. 



Table 1. 





Tno 



Tns 



r„f 





Number of 



Number of 



Number of 



Shrub 



Fruits and 



Fruits and 



Fruits and 





Number of 



Number of 



Number of 





Ovules 



Seeds 



Ovules Failing 



11 



—.093 



.035 



.000 



12 



.131 



.060 



.055 



13 



-.075 



—.090 



.022 



14 



—.053 



— .082 



.103 



15 



— .041 



-.042 



—.009 



16 



—.012 



—.054 



.025 



17 



.173 



—.031 



.159 



18 



.178 



-.001 



.069 



19 



.038 



-.058 



.066 



20 



—.111 



—.056 



-.029 



21 



.010 



-.024 



.025 



22 



—.061 



—.138 



.105 



23 



—.081 



.003 



-.060 



24 



-.017 



-.062 



.006 



25 



.096 



.062 



.005 



26 



-.138 



—.067 



—.044 



27 



—.108 



—.007 



—.075 



28 



.357 



—.035 



.243 



29 



.093 



—.083 



.127 



30 



.097 



-.128 



.163 



For all the pods collected for eacb year, we have: 

 Number per Inflorescence and Ovules per Locule: 

 1906, Table III, y ^ . 0391 + .0086. ;/£'^ = 4.6. 



1908, Table IV. r = . 0633 + . 0061, r/^,. = 10.4. 



1909, Table V, y= — .0539+ .0085, r/£'^ = 6.3. 



Number per Inflorescence and Seeds per Locule: 



1906, Table VI, ;• = — . 0474 + . 0086, r/ Er = 5.5. 



1908, Table VII, ;• = — .0494 + . 0061, rIEr ~ 8.1. 



1909, Table VIII, r = . 0626 + . 0085, r/^,, = 7.4. 



Judg-ed by the probable errors these constants might, taken 

 individually, be regarded as statistically signiiicant. We must note, 

 however, that: 



a) We have taken the number which gives the lowest pos- 

 sible probable error, and that if we used the number of fruits 

 instead of the number of locules, the probable errors would be 

 nearly doubled. 



b) When very high or very low correlation coefflcients are 

 in hand, too great stress cannot be laid upon the probable errors. 



c) The numerical values are too low to be of any practica! 

 biological significance, and positive and negative coefflcients occur 

 for both relationships. 



