438 A CENTURY OF PROGRESS IN THE NATURAL SCIENCES 



obtains when the biological fruit arrangements of the nineteenth century are ap- 

 plied to related taxa. 



B. Seeds: As pointed out by Netolitzky, we actually know so little about seeds 

 in the angiosperms as a whole that it is difficult to use them systematically or 

 to distinguish between structural features which are sufficiently stable to be 

 taxonomically reliable and those which are susceptible to ecological modification. 

 A correlation exists, according to Salisbury (1942), between the amount of re- 

 serve food in the seed and the position the plant normally occupies in plant suc- 

 cession. In some families and genera, seeds have been studied with sufficient 

 thoroughness to permit the establisliment of a basic or ideal "type" for the group, 

 making possible comparisons with those of other groups. Stressing the funda- 

 mental importance in classification of seed structure. Corner (1951) finds that 

 seed morphology strengthens the unity of Leguminosae, and separates the family 

 clearly from Rosaceae. Kratzer (1918) demonstrated the application of com- 

 parisons of seed development to indicate that Cucurbitaceae could not be related 

 to Campanulaceae, Loasaceae, Aristolochiaceae, Begoniaceae, or Ebenaceae; Aris- 

 tolochiaceae and Loasaceae could not be related to Caricaceae and Passiflora- 

 ceae; and Caricaceae could not be related to Euphorbiaceae. Findings from seed 

 characters, he warned, are largely of negative value and cannot be used in dis- 

 regard of evidence from other features of the plant. Reeves and Hutchinson 

 (1947) have successfully employed seed characters taxonomically in Malvaceae, 

 Riek-Haussermann (1944) used them within the genus Veronica, and Buxbaum 

 attempted to relate Cactaceae with Caryophyllales by mutual possession of a 

 type of arillate seed. Murley's (1951) painstaking investigation of seed struc- 

 ture in Cruciferae, beautifully illustrated, exemplifies the kind of information 

 required to permit more general and productive use of seed characters in 

 classification. 



Well-established phylogenetic trends are few and shaky in this area, as should 

 be expected from the dearth of extensive comparative (systematic) data. Aril- 

 late seeds were regarded by Sporne as primitive, because of a positive corre- 

 lation with other characters believed to be primitive, and Corner assigned them 

 a major role in his hypothetical primitive angiosperm. However, Netolitzky 

 tliought all such features as arils and wings to be specializations. The last 

 author stressed that retention by seeds of indehiscent fruits of a well-developed 

 testa is an indication of an ancestral condition, since in such cases the pericarp 

 tends to take over the protective function of the seed coats. Thus, the thinning 

 of seed coats seems to be a general evolutionary advance; the reduction of nutri- 

 tive tissue and of the embryo, or the filling of the latter with reserve foods, are 

 usually regarded as advances, also. Nagaraj (1952) cities the lack of endosperm 

 in the seeds of Salicaceae as an argument for an advanced rather than a primi- 

 tive position for this family. 



The only attempt to develop a comprehensive phylogenetic classification of 

 seeds of which I am aware is that of Martin. The result of an examination of 

 the seeds of more th^n twelve hundred genera, the scheme places primary em- 

 phasis on the shape, size, and position of the embryo. Some confusion was 

 caused by the fact that the author reduced his major divisions from three to 

 two in a terminal footnote of tlie same paper, with the remark: "The resulting 

 two divisions, Peripheral and Axile, seem to represent well the two main lines 



