chosen to represent phyla or classes in cladograms truly representative of the higher taxa
analysed? I would argue that if the species chosen are extant, they cannot possibly be
representative of the character set that existed within the higher taxon 500 million years
ago in the Cambrian. In both cases this is an argument for dealing with species rather than
higher taxa, because we can apply more rigorous analytical tests to data on species.
The conclusion that the succession of species in the fossil record is reliable does not
depend on its completeness. If the fossil record consisted of just two different fossils, the
probability that they would occur in the correct stratigraphic order would still be
overwhelmingly large. If you think about it, this conclusion should not be so surprising. If
the order of species in the fossil record were not reliable, we could not correlate using
fossils. So there really are sound reasons for believing that the order in which species
occur in the fossil record is accurate (irrespective of its completeness). Hence it is
legitimate to use the order of occurrence of fossil species as a test of inferred cladograms.
Detecting genuine ghost rangesHowever, the problem still remains as to how we distinguish between an artificial ghost
range and a genuine gap in the fossil record. Irrespective of cladistics, some ghost ranges
must be genuine. In these cases as more fossil specimens are collected the first known
occurrences of the species should become earlier in time, that is the ghost range should
decrease. This is another example of investigating growth of knowledge or the ‘test of time’.
Paul (1980, 1982) used the idea to estimate the relative completeness of the fossil records
of species, genera, and families of cystoids, and to show that an example from the
neontological record (black smoker faunas) was probably less completely known than the
fossil records of molluscs and echinoderms (Paul 1998). Foote (1997) used this approach
to investigate growth of knowledge of morphological diversity. It has long been used in
the form of species/area curves in plant ecology (e.g. Cain 1938) and is often called the
‘collector curve’. Benton (1998) provided an example for dinosaur species. A curve of the
general shape shown in Figure 5.5 should result. The curve is initially steep, but flattens
out as more specimens are collected. In the case of ghost ranges, early discoveries should
add to fossil ranges, but extending stratigraphic ranges will become progressively more
difficult as more and more specimens are discovered.
MethodData for first records of fossil families are adequate for the present purposes and readily
available in the literature. Initially, I planned to compare the earliest records of families of
fossils using the two versions of ‘The Fossil Record’ (Harland et al. 1967; Benton 1993) to
see what proportion of fossil families had had their stratigraphic ranges extended back in
time over the interval between 1967 and 1993. However, in many major groups there
were significantly more families in the later volume, more than twice as many for teleosts
and mammals, for example. Even with a group such as the graptolites, where the number
of families was almost the same in both publications, there were still differences in the
precise composition of many families due to taxonomic refinements and changes in
classification between the two volumes. Thus the original approach proved impractical
98 CHRISTOPHER R.C.PAUL