84 



Fishery Bulletin 93(1), 1995 



the average frequency of T's (0.29) multiplied by the 

 total number of substitutions (256) divided by three, 

 or 44.37 (Table 2). 



The weights used for each substitution type (Table 

 2) are the ratios of expected substitutions divided by 

 observed substitutions for that substitution type, 

 rounded to the nearest integer (expected divided by 

 observed ratios, or EOR's). All EOR's less than one 

 were rounded to one. Weights were entered into 

 PAUP 3.1. (Swofford, 1991) in the form of a step 

 matrix. 



Results 



Sequence evolution and interfamilial 

 relationships 



Molecular data sets, such as the cytochrome b se- 

 quences presented in this study, are known to encom- 

 pass subsets of characters that evolve at different 

 rates. Subsets of data that differ in their evolution- 

 ary rates will also differ in their phylogenetic utility. 

 Character state changes that accrue very rapidly 

 should permit resolution of very recent divergences. 

 However, these rapid character state changes can 

 provide false inferences about distant relationships 

 because of homoplasy The likelihood of reversals and 

 independent acquisitions is high if a particular site 

 is evolving rapidly because there are only four pos- 



sible character states (G, A, T, and C) and only six 

 possible types of character state change (A<=>G, CoT, 

 GoT, GoC, AoT, and AoC). Therefore, in order to 

 make an accurate reconstruction of the earliest 

 branching events in scombroid history, we should 

 emphasize slowly evolving character state changes. 

 In an effort to best utilize the phylogenetic infor- 

 mation from both slowly and rapidly evolving char- 

 acter state changes, our phylogenetic analysis pro- 

 ceeds in several discrete steps. We begin with an 

 unweighted analysis of all informative nucleotide 

 sites. This analysis is strongly influenced by nucle- 

 otide substitutions that accrue rapidly and should 

 be most informative concerning recent speciation 

 events. We then attempt to improve our resolution 

 of more ancient divergences by giving greater weight 

 to less frequent types of nucleotide substitutions. We 

 conclude with a phylogenetic analysis based on the 

 inferred amino acid sequences. The amino acid se- 

 quences evolve very slowly and should provide our 

 most reliable estimates of the earliest splits between 

 lineages. In each instance, the phylogenetic analy- 

 sis is preceded by a discussion of the evolutionary varia- 

 tion in the character subset under consideration. 



Unweighted nucleotide analysis 



A 590-base pair fragment of the cytochrome b gene, 

 representing positions 134 through 723 of the hu- 

 man cytochrome b sequence, was aligned across all 



