Recent technical advances in the molecular estimation of divergence
times have provided molecular evolutionists with promising tools to
introduce reliable time scales to molecular phylogenetic trees .
One of the most significant advances common to these new methods is the
departure from the molecular clock assumption, which in many cases does
not strictly hold. Another advance is the use of time constraints at
multiple nodes, rather than the assignment of a discrete time value to
a particular node, for rate calibration. This is useful because of the
various uncertainties in divergence time estimations based on fossil
evidence. In general, the occurrence of the earliest fossil assignable
to a particular branch can define the lower boundary of divergence time
for the node at which this branch departed from its sister branch .
However, when the corresponding fossil data are inadequate or sparse,
the lower time boundary based on such data could considerably postdate
the true divergence time, potentially leading to inaccurate or
imprecise dating results [2,3].
In general, fossils of bony fishes are not considered well
preserved. Of the 425 teleostean families, 181 families do not have a
fossil record. Of the remaining 244 that have fossil records, 58 have
only otoliths . Thus, lower boundary values of divergence times based on teleostean fossil evidence could underestimate the true values [5-7].
Therefore, alternative methods that may provide effective time
constraints in dating teleostean divergences should be explored, e.g.,
methods based on reasonable biogeographic assumptions. Because
freshwater fishes do not disperse easily through saltwater, their
evolution may be tightly linked to the geological history of the
landmasses on which they evolved [8,9].
Thus, evaluating the potential correlation of continental drift and
lineage divergences in each of the freshwater fish groups that have
multicontinental distributions is important .
Cichlids (order Perciformes: family Cichlidae) are freshwater fishes
that are mainly distributed in landmasses of Gondwanaland origin
(Africa, South and Central America, Madagascar, and Indo/Sri Lanka) .
They have experienced an explosive radiation in the Great Lakes of East
Africa, and they constitute one of the best-known model organisms for
evolutionary biology .
Phylogenetic studies based on morphological and molecular evidence have
consistently recognized the monophyletic origin of the family, basal
divergences of Malagasy and Indo/Sri Lankan taxa, and the sister-group
relationship of African and South American clades [13-16].
These patterns of divergence among continental cichlid groups are
entirely consistent with the geological history of continental drift,
the proposed Gondwanan origin of Cichlidae, and subsequent vicariant
divergences [5,6,13-18]. However, only a few molecular studies [7,19]
have attempted to evaluate this hypothesis by dating cichlid
divergences; their different approaches led to opposite conclusions.
Genner et al.  supported vicariant cichlid divergences during Cretaceous times (vicariant hypothesis), whereas Vences et al. 
suggested a Cenozoic transmarine dispersal (dispersal hypothesis). The
latter conclusion is more consistent with the Eocene occurrence of the
oldest cichlid fossils .
We used molecular data obtained from complete mitochondrial DNA
(mtDNA) sequences to investigate these hypotheses. Among the 54 fish
taxa that we sampled, we newly determined the sequence data for six
cichlid species. The two alternate hypotheses for cichlids, vicariant
and dispersal ones, were evaluated by estimating the divergence times
of the taxa using Bayesian analyses that incorporated extensive
fossil-based time constraints for various divergences. Despite the
relative paucity of fish fossil records, this set of time constraints
allowed us to estimate cichlid divergence times with high enough
resolution to discriminate between the two alternative hypotheses.