Page 9 - iCC044-201501_March

places this continental fragmentation event
in the Early Cretaceous (ca 135 Ma) [11].
However, the stratigraphically oldest fossil
cichlids are Eocene in age (approx. 46 Ma)
[12,13], implying a gap of approximately 90
Myr in the early history of the group. This,
along with the absence of Early Cretaceous
fossils belonging to more inclusive and
taxonomically diverse clades that con-
tain cichlids, has led some to abandon
the orthodoxy of Mesozoic vicariance in
favour of Cenozoic dispersal to explain the
present-day distribution of cichlids [12,14]. non-evidence; they are hypotheses
The complete absence of fossil cichlids
from many former Gondwanan landmasses
would seem equally problematic for the
vicariance hypothesis, but has received
surprisingly little attention. For example,
the Australian fossil record contains sev-
eral fish-bearing freshwater deposits of
Mesozoic and Cenozoic age, but no fossil
cichlid is known from the continent. While
it is clear that assembly of the composi-
tionally distinctive Australian freshwater
fish fauna has a complex history stemming
from isolation, aridification and marine
invasion coupled with the persistence of
ancient lineages [15], this complexity does
not undermine the prediction of the vicar-
iance model that cichlids should have
been widely distributed across Gondwanan
landmasses during the Mesozoic [9].
These palaeontological arguments have
been dismissed as ‘non-evidence’ by
advocates of cichlid vicariance [16]. Some
authors have even suggested that the
fossil record supports the notion of cich-
lids deep within the Mesozoic, citing the
high probability of non-preservation for
freshwater taxa of Cretaceous age [9]
cichlidpower.org.au
or inferring that the advanced morphol-
ogy of the earliest fossil cichlids implies
a long, and as yet unsampled, palaeon-
tological history of the group [9,10,13].
The seemingly ambiguous signal of
palaeontological data with respect to
the question of cichlid origin is symp-
tomatic of a qualitative approach to
an inherently quantitative problem.
Invoked stratigraphic gaps are neither
ad hoc contrivances nor trivial incon-
veniences to be dismissed as
amenable to statistical interrogation.
In order to provide a robust time scale for
cichlid diversification and select between
competing biogeographic hypotheses, we
applied three semi-independent approach-
es in estimating the age of crown-group
Cichlidae. Our first two methods are pal-
aeontological, and draw on (i) the distribu-
tion of fossil horizons yielding cichlids and
those that might plausibly yield cichlids (i.e. the sampled range of the group of inter-
fish-bearing freshwater deposits on former
Gondwanan landmasses) [17], and (ii) the
stratigraphic distribution of more inclusive
teleost lineages (meaning clades of higher
taxonomic rank) that contain cichlids [18].
These techniques relate directly to two
contrasting arguments that emerge repeat-
edly in palaeontological debates concern-
ing the chronology of cichlid evolution:
either that the record of freshwater fishes
generally, and cichlids specifically, is suf-
ficiently poor that the absence of Mesozoic
cichlid fossils is unsurprising, or that the
minimum age of origin for a series of more
inclusive lineages of teleost fishes pre-
cludes the origin of cichlids deep within
the Mesozoic. Significantly, these methods
share only one common feature in their
calculations: both are necessarily con-
strained by the minimum age for cichlids as
imposed by the oldest fossil example(s) of
the group. As an independent assessment
of the divergence times estimated from
palaeontological data, we conducted a
relaxed-molecular-clock analysis for cich-
lids and Ovalentaria [19], a percomorph
lineage that includes cichlids. Ourdataset
includes 10 protein-coding nuclear genes
for 89 cichlids and 69 non-cichlid species
of Percomorpha.
2. Material and methods
(a) Estimating time of evolutionary origin using the
distribution of cichlid-bearing fossil horizons
One method of estimating credible inter-
vals (CIs) on stratigraphic durations draws
on the number of fossil horizons within
est. The simplest approach assumes that
fossil horizons are distributed at random
[20,21], but the potential for fossil recovery
undoubtedly varies over time as a conse-
quence of a heterogeneous rock record.
Marshall [17] developed a more general
method that permits non-uniform preser-
vation by using an empirically informed
function that quantifies potential for fos-
sil recovery. We have applied this logic
in conjunction with a Bayesian approach
that provides a statistically appropriate
framework for discussing the probability
of clade origin within certain stratigraphic
intervals [20]. Our results are conditioned
on the prior assumption that cichlids are
post-Palaeozoic in age (i.e. they originated
in the Triassic or later), which is consistent
with the fossil record and does not exclude
the possibility of Gondwanan vicariance.
7