ttwttwtthw*^^^ 



FIGURE 20. — Additive binary matrix based on relationships of 

 Unicolax parasites. 



to 



I 



1 



I 



5 



-C 





c 



c 



•20- 



U L,J 

 L_„_l 



L 



LJ 



J 



UJ 



,J 



.19. 



■21- 

 I 



FIGURE 21. — Cladogram of scombrid hosts based on host morphol- 

 ogy. Nodes 12-21 represent hypothetical ancestors. 



set of the Scombridae. The additive binary matrix of 

 this tree is presented in Figure 22. 



In Figure 23 we have indicated the scombrid genera 

 in the tribes Scomberomorini, Sardini, and Thunnini 

 parasitized by Unicolax, based on the phylogeny of 

 the Scombrinae proposed by Collette and Russo 

 (1979). The copepod species are ranked from the 

 most plesiomorphic (generalized) to the most apo- 

 morphic (specialized), based on the Wagner tree of 

 Unicolax (Fig. 19). 



As stated earlier, parasite phylogenies can be coded 

 as characters and used to generate host trees; con- 

 versely, host phylogenies can be coded as characters 

 and used to generate parasite trees (Brooks 1981). In 



cases where a host has more than one parasite or a 

 parasite has more than one host the character states 

 for the two series are inclusively OR'd (Copi 1972) 

 and a single series is used. By logically OR'ing two 

 characters, a character state is said to be present in 

 the union of two groups, if and only if it is present in 

 one or both groups. For example, in Figure 20, Auxis 

 harbors U. collaterals (2) and U. mycterobius (4). The 

 character states for a host bearing U. collateralis can 

 be determined by reading across line 2 of the additive 

 character matrix, that is a one or logical true for states 

 2, 8, and 9 and not true for the others. The character 

 states for a host bearing U. mycterobius can be deter- 

 mined by reading across line 4, that is a one or logical 

 true for states 4, 6, 7, 8, and 9 and not true for the 

 others. Logically OR'ing the two rows of the matrix 

 results in the character states 2, 4, 6, 7, 8, and 9 being 

 true and the others being not true. Referring to the 

 parasite tree (Fig. 19), these character states repre- 

 sent the host, A axis, as having or having had during 

 the course of its evolution (sensu lato) parasitic taxa 



(2) U. collateralis, (4) U. mycterobius, and hypotheti- 

 cal ancestors (1), (2), (3), and (4). 



Proceeding in this manner for each host, a parasite 

 (parasite ancestor) by host matrix is constructed. 

 This matrix was subjected to cladistic character anal- 

 ysis using the WAGNER 78 program for optimiza- 

 tion. The resulting Wagner tree (Fig. 2 4) is rooted at a 

 hypothetical host ancestor without Unicolax parasites. 

 According to Brooks' (1981) methodology, this tree 

 is an estimate of host phylogeny in lieu of host mor- 

 phological data. It estimates host phylogeny based 

 on phylogenetic events of their parasites. Because we 

 have a host phylogeny based on morphological data, 

 a direct comparison between the two trees is pos- 

 sible. We attempt to explain the source of differences 

 between the estimate of host phylogeny based on 

 parasites and a cladogram based on host morphol- 

 ogy. 



The most notable difference is that the base, node 

 (5) of the host by parasite tree (Fig. 24), is formed by 

 an unresolved multicotomy. This has resulted be- 

 cause it is more parsimonious to assume that the four 

 scombrid taxa, which lack Unicolax, never had them 

 than to assume they were first acquired then lost. 

 Node (4) is a subset of node (21) on the host phy- 

 logeny (Fig. 21) and is based on a common Unicolax 

 ancestor [node 1, (Fig. 19)]. Node (3) is a subset of 

 node (19) on the host phylogeny and is based on the 

 presence of ancestor (2) and parasite (2), U col- 

 lateralis. An unresolved tricotomy is present at node 



(3) because the only parasite shared by the hosts 

 Cybiosarda and Orcynopsis is U. collateralis, which is 

 present below node (3) and is therefore treated as 



256 



