1. The present paper represents the final report of the experiment about the constancy in a clon of the orange "Baianinha" which has been studied during the past four years. 2. The trees were obtained by grafting buds taken from one tree, 17-2, of the "Asilo" in Piracicaba, on non-selected seedling root stock of sour orange. 3. The fruits were collected separately for each tree and their diameter (in mm) determined. The test "within-between" showed clearly that the variation between the means of trees and within trees are of a totally different order, the former being about 10 times bigger than the latter. 4. Thus it was necessary to use other tests in order to be able to find out whether the trees form a uniform sample or not. A X²-test was carried out for the mean diameters of the 217 trees included in the experiment in 1937 and demonstrated the existence of an accentuated heterogeneity. The same method however could not be applied in the analysis of later years in view of the reduced number of trees. It was necessary therefore to device another test for the data of these years with 44 or 43 trees each only. In carrying out this "test of seguence" the mean diameters were arranged according to their magnitude and were given accordingly numbers of order. The probability was then calculated with which a tree may obtain by chance and in successive years such numbers of order as registered (quadro V). Fuller details are given in the test of this paper and shall be published elsewhere as well. The results of this test were in accordance with those obtained with the X²-test. 5. The trees do not form a uniform sample as explained above. Two types of heterogeneity exist: a) The variation between trees is of an entirely different order as the variation within trees, with of 7,2 to 16,5 (guadro III). It may be remembered that ft (Brieger) is identical with Vf (Snedecor) and e z (Fisher). b) The variation between trees is highly heterogeneous as shown by the x2-test (guadro IV) and the "test of sequence" (quadro V). It was possible to separate the sample into three groups : the type "Baianinha Piracicaba" with a medium mean fruit diameter per tree, the "Baía" with large fruits and the new type "Baianinha Araras" with small fruits. 6. The experimental plot was selected as beeing reasonably uniform. The statistical analysis however revealed significant, though not very important heterogeneities which can only be partially responsible for the variations between trees mentioned above (quadro VI). 7. There is a constant negative linear correlation between the number of fruits per tree and the mean fruit diameter, the coefficient r beeing the same in all years and equal to 0.5 (quadro X). On the average, a loss of 1 mm in mean diameter corresponds to a gain of 50 fruits per tree. 8. The error within trees both when comparing the errors per tree in the same year or of different years showed significant variations which are however without much praticai importance. The annual means of these errors (sD) vary from 4,69 mm in 1937 to 7,56 mm in 1939 (quadro VIII). 9. The error within trees cannot be used as a reliable measure since the variations within trees are highly abnormal, restricted by two physiological limits which prevent the appearance of very small or very large fruits and show sometimes a tendency to asymetry and bimodality. 10. The possible causes for the abnormal variation "within" were studied in some detail. The only positive result was obtained when comparing the production on older branches with that on recent strong off-shoots. In some case it was possible to collect the fruits on these components of a tree separately, the older branches forming the lower and the other the top parts of the tree (fig. 7). The latter produce always larger fruits (quadro IX and fig. 2). 11. Heterogeneity of the root stocks seems to be the most probable cause of the heterogeneity between trees. It has been shown by other experiments, not yet published, that all species used as root stocks in this country are highly variable which is not astonishing since there has not yet been any selection for root stocks (fig. 8a to 10b). A-similar variation has been found by Webber for the Californian sour orange. 12. The existence of effects of root stocks on scion, besides the appearance of general incompatibilities, has been shown in the extensive studies of Hatton. Webber (12) demonstrated correlations between root stock and the trunk area, top volume and total yield of scion in Washington Navel budded on sour orange in California. Gurgel (3) found an influence of various root stocks on pollen fertility in the "Pera" orange and the grape fruit "March seedless". Further experiments, already under way, will show if our hypothesis about the effect of the root stock on number and diameter of the fruits of the scion is correct. 13. The variation of seedlings in Citrus cannot be due exclusively to the occurrence of sexual reproduction besides simple apogamy as supposed by Webber. This author assumed that all variants were due to the appearance of embryos from normally fertilized eggs. However histological complications seem to occur which cause variations also within the apogamous offspring. 14. The praticai value of the new type "Araras" in contrast to "Piracicaba" is illustrated in figures 4 and 5 and quadro XIII. The proportion of larger fruits of the commercial types of "126" to "80" is very small in "Araras". Under the pre-war conditions the type "126" was still exportable without loss while the larger fruits could not be exported at all. Thus it would be of great economic value to establish definitely the new type "Baianinha Araras".