How Many Dinosaurs Were There? Counting Species Nobody Can Interbreed

The most widely taught definition holds that a species is a group of populations that interbreed or could interbreed and are reproductively isolated from others, which is a useful criterion for living sexually reproducing organisms and is untestable for anything known only from bones. Palaeontologists therefore work with a morphological criterion, treating specimens as one species if they fall within the range of variation expected within a population and as separate species if the differences exceed it. That requires knowing what range to expect, which requires a sample large enough to measure variation, which is exactly what most fossil species lack. The result is that the decision rests on judgement informed by comparison with living relatives, and reasonable people reach different conclusions from the same specimens.

Several things produce variation between specimens and each can be mistaken for a species difference:

• •Growth, since juveniles differ from adults substantially and in ways that change with age
• •Sex, where males and females differ, which is common in living animals and difficult to establish in fossils
• •Individual variation, which is substantial in any population and is frequently underestimated from small samples
• •Geographic variation within a species, which grades continuously and has no natural cutting point
• •Distortion during burial, which alters proportions and can be mistaken for anatomy
• •Genuine species differences, which is what everything else has to be excluded to establish

The field has long contained a split between workers who name many species from small differences and those who group specimens into fewer variable species, and the labels for the two tendencies are old and slightly unkind. The first approach risks naming variants and growth stages as species, which inflates diversity counts and clutters the literature with names that later have to be retired. The second risks lumping genuinely distinct animals together, which conceals real diversity and makes evolutionary patterns harder to see. Neither is a methodological error as such, since the underlying question has no determinate answer, and the difference frequently reflects how much variation the worker believes a single fossil species should contain. Statistical approaches comparing fossil variation against measured variation in living populations have narrowed some disputes without settling the general question.

Several long-running disputes have been resolved and the methods that resolved them are the useful part. Bone sectioning establishes maturity independently of size, since a specimen still growing rapidly cannot be an adult of a small species, and applying that to disputed pairs has shown some supposed species to be juveniles. Large samples from single sites allow variation to be measured directly rather than assumed, and several sites have produced enough individuals of one kind to establish what a population actually looks like. Statistical tests comparing the pattern of variation in a fossil sample against samples from living species distinguish a single variable population from two overlapping ones in some cases. And growth series with intermediate specimens settle the matter outright when they are found, which is why finding juveniles matters far more than the modest attention such specimens usually attract.

The number of species present is not a bookkeeping matter, since a great deal is calculated from it. Diversity through time is the basic measure of how the group was faring, and whether dinosaurs were declining before the asteroid impact has been argued for decades using counts that depend entirely on these decisions. Geographic patterns depend on whether populations in different regions are counted as one species or several. Rates of evolution depend on how many species existed over a given interval. And the rock record adds its own problem, since more species are found where more rock of the right age is exposed and more collecting has happened, so apparent diversity tracks sampling effort as well as biology, and correcting for that is a substantial research area in its own right. Published diversity curves therefore carry two large uncertainties at once.

The interbreeding criterion cannot be applied to fossils, so specimens are grouped by whether differences exceed the variation expected within a population, which requires a sample most species lack. Growth, sex, individual and geographic variation and burial distortion all mimic species differences. Diversity counts, including whether dinosaurs were declining before the impact, rest on these judgements.