The known fossil record contains roughly a quarter of a million species, most ofwhich are extinct. Although fossils of the earliest forms are important to our knowledge of the history of life, the fossil record is dominated numerically by the remains of multicellular organisms from the last 600 million years (Myr). Fossil species are grouped into about 35,000 genera and 4000 families. About one-quarter of the families are still living.
Although the fossil record is ample for statistical purposes, it contains a very small fraction of all the species that have ever lived. Estimates of that fraction range from
The dinosaur fossil record illustrates some of the more extreme sampling problems. According to a review by Dodson (7), 336 of the named species of dinosaur are taxonomically valid. Of these, 50%o are known only from a single specimen, and about 80%o are based on incomplete skeletons. The 336 species are grouped into 285 genera, and of these, 72% have been found only in the rock formation where they were first discovered, and 78% have been found in only one country. These numbers are astonishing if viewed as if the data were complete. The species/genus ratio being barely above unity is undoubtedly due to incomplete sampling, as is the apparent biogeographic restriction.
Incomplete sampling also influences our estimates of extinction rates. Lack of fossilization inevitably shortens the apparent life span of species, and this may explain why durations of dinosaur species are far shorter than is typical of other, better preserved organisms. On the other hand, shortlived, localized species have a low probability of appearing in the fossil record at all. The net effect of these biases is that statistical estimates of mean duration are almost certainly exaggerated. That is, the fossil record is biased in favor of successful species-successful in the sense of surviving for a long time and being ecologically and geographically widespread. Thus, analysis of past extinctions must operate in a sampling regime very different from that of present-day biodiversity.
Fig. 1 shows the frequency distribution of recorded life spans of 17,500 genera, of fossil marine animals. The distribution is highly skewed, with the mean (28 Myr) being the result of many short durations combining with a few very long ones. Survivorship analysis of the genus data indicates a mean species duration of 4 Myr (9), although as indicated above, this is probably a high estimate because of the dominance of successful species in the sample. Regardless of the uncertainties, however, species and genus residence times on earth are very short on geologic time scales. The longest-lived genus in Fig. 1 (160 Myr) lasted only about 5% of the history of life.