1912] HARRIS—STAPHYLEA 397 
lation for length and seed/ovule index and the partial correlation 
coefficient. Comparing 7, and 7; for the combined series we have: 
Relationship 1906—2050 pods | 1907—1218 pods 
Number of seeds and length, oY ioe. 0.3522+0.0131 | ©. 2019+0.0185 
ae index and length, r 4 ©.3418+0.0132 0. 2018+0.0185 
Difference, Asad caer a poe 0.0104+0.0186 | 0.0001 +0.0261 
Both differences are less than their probable errors, me of no 
significance. 
Consider now the correlation between s and / for constant values 
of 0, as measured by the partial correlation coefficient of pis. The 
correlations for number of ovules formed and number of seeds 
developing per locule (r,,) are necessary. The tables of data are 
36 in number, and since they are supplementary rather than funda- 
mental to our main subject, we need not publish them. The con- 
stants with their probable errors are set forth in tables XI and XII. 
TABLE XI TABLE XII 
Number of os Number of os l | E 
shrub eae ioe crales r/Er ahha agiee gun Pg wks | "/Er 
| Daa —0.0760.037 | 2.06 oh estilo 0.156+0.060 | 2.60 
LS Peer 0.0420.035 1.19 | aoeceae 0.021+0.054 | 0.39 
| ae 0.1540.037 | 4.16 Ge. cs —0.0720.046 1.56 
(FS esheets O.112+0.038 2.96 PP ea 0.159+0.038 | 4.23 
Ener —0.131+0.037 | 3.54 oS 0.0280.047 | 0.60 
bE Rees 0.063+0.039 t.61 BE is 0:057+0.038 | 1.50 
Le marae eae 0.080+0,.039 2.05 ae 0.094=0.043 | 2.18 
ee ©.002*0.039 | 0.05 cc pape th 0.126*0.126 | 3.19 
ce a —0.038+0.039 | 0.96 Pras 0.0250.025 0.56 
ee 0.045 +0.037 1.20 Secs .0.079+0.079 | 1.93 
4 Oe —9.094+0.039 2.40 BO soe 0.150*0.049 | 3. 
RR ©.028+0.039 0.71 1b AE 0.0040. ! 0.10 
5 ren 0.0140.039 0.36 a8. 0.0330.040 | 0.82 
ues 144+0.038 3.78 Poona 0.137=0.05 | 2.33 
5 eee i 0.078+0.039 2.00 pa Hae tha 0.045 39 1.16 
og ne ee ©.026+0.039 | 0.67 ery ©.010#0.050 | 0.19 
se ee 0.048+0.039 1.22 
Webco | -0930.038 | 2.45 
Oe oa ©.058+0.039 | 1.48 
SORA 0.042+0.039 £.08 
i ee cs 
ssibly we are not quite justified i in using the ordinary method of calculating 
the aches error of at is, th f the sum of the squares 
of the two probable errors, but the differences in the correlation are so very small that 
it makes no practical difference 
