The maximum-likelihood coalescent approach, as
implemented in the program FLUCTUATE v.1.4 available
at http://evolution.genetics.washington.edu/lamarc/fluctuate.
html, was used to estimate population growth rate
and effective size. FLUCTUATE utilizes a model of a single
population that has been expanding (or declining)
exponentially and provides estimates of the parameters
Q and g. Q is defined as 4n1l where n1 is the ‘current’
effective population size, l is the neutral mutation rate
per site; g is the exponential growth rate of the population
and is positive if the population is expanding and
negative if the population is declining. We used Watterson’s
(1975) segregating sites estimate to obtain the
initial estimate of Q for each run, and a transition/
transversion ratio of eight based on estimates from
MODELTEST 3.06 (Posada and Crandall 1998). One hundred
short chains (1,000 steps per chain with a sampling
increment of 100) followed by 10 long chains (500,000
steps per chain with a sampling increment of 500) were
employed to ensure convergence of the Markov Chain
Monte Carlo. Estimates of time (in generations) since a
‘population’ differed from the current effective size were
determined by using the formula ts=(ln(nt/n1))/(gl),
where n1 and nt are the effective population sizes at
present (n1) and t generations (nt) ago, respectively.
Estimates of l employed were 1.0 and 1.5% per MY and
were based on molecular-clock calibrations of the
mitochondrial COI (1.2%/MY1) and NADH-2 and
ATPase-6 (1.3%/MY1) genes developed for several
geminate species pairs of fishes (Bermingham et al.
1997).