Cooper and Chapleau: Monophyly and intrarelationships of the family Pleuronectidae 
691 
(Table 1), illustrating the distribution of 106 charac- 
ters for 58 taxa, combines information from the lit- 
erature (Norman, 1934; Batts, 1964; Amaoka, 1969; 
Ahls trorn et ah, 1984; Hensley and Ahlstrom, 1984; 
Sakamoto, 1984a) as well as 67 new morphological 
features observed through examination of cleared 
and stained material, whole preserved specimens, 
and radiographs. Morphological states obtained from 
the literature are indicated by the citation immedi- 
ately following the italic character description (Ap- 
pendix). Morphological states based on meristic 
counts represented the modal value of the sample 
population. Characters at each node of the cladogram 
are described and numbered in order of presenta- 
tion in the text. 
Given the unresolved nature of the interrelation- 
ships for Pleuronectidae and other bothoid taxa 
within the order (Chapleau, 1993), the establishment 
of character polarity by outgroup comparison, as 
outlined by Watrous and Wheeler (1981) and 
Maddison et al. (1984), was not possible. Character 
polarity was determined through a direct examina- 
tion of states observed in outgroup taxa (Table 1). 
For each character, the majority state observed in 
the three bothoid taxa Citharichthys arenaceus, 
Paralichthys lethostigmus, and P. squamilentus was 
assumed to represent the plesiomorphic condition. 
This decision was only overruled if there was het- 
erogeneity in the distribution of states within these 
three taxa and if both secondary outgroup taxa 
Lepidoblepharon ophthalmolepis and Psettodes sp. 
possessed the alternative state. One exception to this 
rule is stated in character 82 (Appendix). Character 
states hypothesized as plesiomorphic for the family 
are coded as zero (0). 
Heuristic search methods 
The matrix was analyzed with all combinations of 
heuristic search parameters available in PAUP 
3.1. 1. 1 An exhaustive search of the most parsimoni- 
ous tree with this many taxa (53) would have re- 
quired the analysis of an estimated 2.84 x TO 82 bifur- 
cating trees (Felsenstein, 1978) and is not possible 
within the current standard of computational time. 
For these same reasons, a branch and bound search 
technique proved inadequate to resolve relationships 
for more than 25 taxa (Forey et al., 1992). A two-step 
procedure (a sequential addition of taxa to produce 
a cladogram that minimizes homoplasy followed by 
the subsequent branch-swapping of this addition tree 
to search for more parsimonious cladograms) was 
1 Swofford, D. L. 1991. PAUP Version 3.1.1. Unpublished 
software documentation. 
used to search for the most parsimonious result. The 
random addition sequence in combination with all 
of the branch-swapping algorithms was a nonrigorous 
means of assessing the efficiency of the heuristic 
methods (Forey et al., 1992). If 50 random replicates 
give the same set of tree topologies, then it is likely 
that the maximally parsimonious trees have been 
found. However, if after 100 replicates shorter cla- 
dograms are still being found, then it is likely that 
more trees remain (Forey et al., 1992). 
To minimize confounding effects of local minima, 
the “keep” option was used on successive searches to 
allow swapping on nonminimal trees (Forey et. al., 
1992). Both the “MULPARS” and “swap on all trees” 
options were employed during each heuristic search 
to minimize the effect of plateau (Forey et. al., 1992). 
All searches assumed that the ingroup was mono- 
phyletic, and all uninformative characters were ig- 
nored. Character optimization was set for acceler- 
ated transformation ( ACCTRAN ). The five outgroup 
taxa were not included in the analysis. Instead, an- 
cestral states for all characters were set as zero ac- 
cording to established character polarity to repre- 
sent a hypothetical outgroup. All most parsimonious 
trees were saved from each search and combined 
(without duplication) to establish a 50% majority- 
rule consensus of the equally parsimonious results. 
Character analysis, character consistency index (cci), 
tree statistics, and tree presentations were gener- 
ated with MacClade version 3.04 (Maddison and 
Maddison, 1992). 
Results and discussion 
Phylogenetic analysis 
The heuristic searches found multiple trees of equal 
length (Table 2). The most parsimonious trees were 
found to be 403 steps from a minimum of 131 steps, 
with a consistency index (ci) of 0.33, excluding unin- 
formative characters, and a retention index (ri) of 
0.79. Additional rounds of heuristic search allowed 
swapping on nonminimal trees up to 410 steps but 
did not resolve cladograms shorter than 403 steps. 
The “simple” and “closest” addition sequences were 
biased by taxa with large numbers of unknown char- 
acter states. Criteria for establishing the initial tree 
and the addition of taxa are strongly influenced by 
unknown character states in these two algorithms. 
As a result, Reinhardtius evermanni, Clidoderma 
asperrimum, Hippoglossoides dubius, Microstomus 
shuntoui, Pleuronichthys coenosus, and Pseudo- 
pleuronectes obscurus were not included in the heu- 
ristic searches. The “as is” and “simple” addition se- 
