FUIMAN: OSTARIOPHYSI 



135 



Table 28. Character-State Changes on Stems Leading to 

 Hypothetical Ancestors (Nodes) and Terminal Taxa on the 

 Wagner Tree of Ostariophysi. Numbered character states correspond 

 to those given in Table 27 , Uniquely derived, unreversed character states 



are given in boldface type. Reversed characters are noted by (r). Node 

 numbers correspond to those given in Fig. 67. 



Node Characler state 



1 8c, lib, 12b, 14b, 20b, 20c 24b 



2 6b 



3 14d, 15b, 18b 



4 13a, 21a 



5 18a(r), 20b(r) 



6 16b 



7 17b 



8 14c, 20a(r), 23b 



9 6b, 19a. 24a(r) 



10 3b, lla(r), I2a(r). 14b 



11 lla(r), 22b 



12 6b, 6c 



13 3b, 4c, 12b(r), 14a, 14b, 14c 



14 3a, 4b, 18a(r), 20b(r) 



15 10b, lib, 24a(r) 



16 12b. 17b. 14b(r) 



17 2b, 6b, 7a. 8b(r) 



18 8a, 16b. 20b{r) 



19 12b, 16a, 18a(r), 20a(r) 



20 2a(r), 6a(r). lOa(r) 



2 1 5b, 6c, 7b(r), 8b(r), 9b, 1 1 c, 1 5c, 17a, 1 7b. 1 9a 



22 lOa(r), 15c, 15d 



23 14b(r) 



24 lb, 2c, I7d 



25 6a(r), 1 la(r), 15c(r), 16d, 19c 



26 Ic, 15e, 19a, 20b(r). 22a 



27 2d, 8a, 14b(r), 14c(r), 17b. 17c(r). 18a(r) 



28 4a, 4b, 6c, 10b, 15f 



by measurable gaps were coded individually. Characters were 

 polarized by outgroup comparison (Table 27). The evolutional^ 

 transformation series for each continuous, multiple state char- 

 acter was assumed to be linear (i.e.. with one or two adjacent 

 states for a given state). Consequently, a character coded with 

 n states had n - 1 different changes from one state to another, 

 disregarding the direction of change. These transitions were 

 termed "two-state factors." All two-state factors and their states 

 for each species were generated by the FACTOR computing 

 program (Estabrook et al., 1976). The output from this program 

 included an input file for the WAGNER 78 computing program 

 which was used to construct Wagner trees. The data deck was 

 resequenced and a new Wagner tree generated several times in 

 order to identify the shortest (most parsimonious) tree (Jensen, 

 1981). 



Characters 

 Morphomelhc characters.— To develop morphometric charac- 

 ters for phylogenetic analysis, the following lengths were mea- 

 sured along the longitudinal axis of the fish; total length, preanal 

 length, head length, and eye diameter. Two vertical measure- 

 ments, head depth and body depth at anus, were meant to rep- 

 resent size and shape in the dorso-ventral direction. All mea- 

 surements were defined by Fuiman (1979). They were made 

 reasonably independent of one another by subtracting preanal 

 length from total length to yield peduncle length, and head length 

 from preanal length to yield tnank length. Peduncle length, trunk 

 length, head length, eye diameter, body depth, and head depth 

 comprised the basic morphometric characters. 



(0 



o 







a 

 CO 



0) 



E 



3 



3- 



2- 



1- 



\ZZ} Cypriniformes 



I I Characiformes 



^H Siluroidei 



rXI Gymnotoidei 



0.25 



0.45 



0.65 



0.85 



Yolk-Sac Shape (depth/length) 



Fig. 68. Frequency distribution of yolk-sac shape for recently hatched 

 ostariophysan species. 



Body dimensions of larvae are strongly influenced by allom- 

 etry (Fuiman. 1983b). Such measures cannot be expressed as 

 simple proportions, because the proportions are not constant 

 within a species throughout the larval period. The effect of size 

 on shape must be eliminated in comparisons of shape. Further, 

 any single measure which accounts for size in one taxon may 

 be an inappropriate measure of size in a distantly related taxon. 

 Within-group principal component analysis can be used to ex- 

 tract a size component, PCI (Humphries et al., 1981), that is a 

 linear combination of several variables, each containing infor- 

 mation on size and shape. Thus, PC 1 includes more information 

 on size than any single measure and is a better comparison across 

 taxa. 



Univariate and multivariate methods of allometry relate dis- 

 tance measures log-linearly (Huxley, 1932; Jolicoer, 1963). Thus, 

 a within-species principal component analysis of the logarithms 

 of the six basic morphometric characters, based on the covari- 

 ance matrix, was performed to extract the size component (PCI ). 

 The extreme PCI scores for all taxa were compared and two 

 values (0.00 and 0.60), one near each end of the larval period, 

 were chosen as standard sizes for comparing morphometry. The 

 six morphometric measures were reconstructed for each of these 

 sizes by means of the regressions of the logarithm of the char- 

 acter on PC 1 . By selecting two sizes to compare, the phylogenetic 

 analysis included information on changing shape (allometry) as 

 well as static shape. The final 1 2 character values were recorded 

 as predicted lengths (in mm) for each morphometric measure 

 at each of 2 standard sizes. However, body depth at the anus 

 contained no discontinuous, phyletic variability. The final mor- 

 phometric characters were; (Characters 1 and 2) Peduncle length 

 (smaller and larger standard size, respectively), (3 and 4) Trunk 

 length, (5 and 6) Head length, (7 and 8) Eye length, (9 and 10) 

 Head depth. Three additional morphometric characters were 



