Species in the primate fossil record



Species in the fossil record are population pools of genetic and phenetic variation at a place and time, morphologically recognizable and distinguishable from others by empirical standards. Change through time can be substantial, requiring subdivision of lineages that becomes more arbitrary as they become more complete. Evolution is about form, space, and time; it is about variation and change. Interpretation of species in the fossil record touches all of these.

The fossil record of primate evolution is an embarrassment of riches: there are too many fossils to count, from hundreds to thousands of localities and levels, on five of the northern and southern continents. The record is embarrassing too because of the difficulty scholars seem to have making sense of the profusion.

Here I outline some generalizations that are helpful when interpreting fossils. First, form, space, and time are different. It makes sense to deal with these one by one, starting with form, or morphology (size and shape), at one place and time. Then we can consider form and space together; form and time together; and finally, in theory, form, space, and time simultaneously (a synthesis still to be achieved because of the rarity of well-studied fossiliferous sections and the difficulty of correlating these precisely in time).

Much of the primate fossil record is teeth and jaws because teeth are mineralized and effectively fossilized during life. Teeth have other advantages: the morphology of primate teeth is complex and related to diet;[1] The size of teeth is determined genetically, with high heritability;[2] teeth form and reach definitive size before eruption; and the size of teeth in primates is highly correlated with body size.[3] Most mammals in a fauna can be identified by the size and shape of their teeth, and any error of overlapping similarity will be conservative in underestimating species richness.

The primate fossil record includes many skulls and skeletons in addition to teeth. The guidelines for interpreting variation outlined here apply equally to skulls and skeletons. It is helpful to keep in mind that evolution by natural selection is fundamentally about population variation and change through time. Darwin passed Lamarck studying evolution when he shifted the subject from individuals to populations and population variation. We are not studying immutable archetypes.


Humans are the primate species we know best. We live in a population composed of individuals, each of which is different. We see enough individuals in a day to recognize that we are united by our similarity and distinguished from other species by our differences. Species exist in relation to others. Even in a forest with chimpanzees, we can recognize who is human — because we are similar, and they are different. Behavior shows that we interbreed and they interbreed, but we don't interbreed with them. Reduced to what might represent us as fossils, jaws and teeth, humans are still morphologically similar and chimpanzees are different. Similarity and difference can be studied by measuring morphology, quantifying variation, and expressing similarity and difference in units of variation.

Variation has rules. Biological variation is bell-shaped or “normal.” Normal distributions result from flipping coins or rolling dice, recording the permutations, and tallying combinations. Biological variation of interest in evolution is polygenic in origin, resulting from the additive contributions of many genes, each of small effect. The “location” of a normal distribution is given by its mean and the dispersion of a normal distribution is given by its standard deviation. If we know the standard deviation, then we know the expected range, because 95% of the variation will lie within ±2 standard deviations of the mean and 99% will lie within ±3 standard deviations.

We measure arithmetically by counting units of standard size, but functional relationships are geometrical and proportional. Logarithms provide the required transformation. Empirically, biological variation is log-normally distributed within species,[4] variation is standardized relative to the mean when measurements are logged, and allometric relationships between species are linear when logged (allometry is another word for geometry). The natural logarithm (ln) is the transformation of choice for converting measurements to a proportional scale because the standard deviation on a natural-log scale is mathematically the coefficient of variation (standard deviation divided by the mean) on an arithmetic scale.[5]

Coefficients of variation for linear measurements of tooth size, stature, and so forth within a species generally average about 0.04-0.05.[4],[6-10] Areal measurements of tooth size average about 0.08-0.107,[9, 10] and volumetric measurements of body weight or brain weight average about 0.10-0.15.[4, 11] These are related to the dimension of the measurements. Thus, measurements of length for a species can be expected a priori to have a range of about 0.20 units on a natural log scale, measurements of area can be expected to have a range of about 0.40 units on a natural log scale, and measurements of volume or weight can be expected to have a range of about 0.60 units on a natural log scale.

The standards work both ways. When converted to natural logarithms, molar lengths with a range greater than 0.2 are likely to represent more than one species but, all else being equal, a range of less than 0.2 probably represents a single species. The same holds for molar crown areas for a range of 0.4 units, and for brain volumes for a range of 0.6 units. It is important to consider distributions as well as ranges, but range standards provide a guide for the interpretation of samples large and small.

Figure 1 illustrates how logarithms bring simplicity and regularity, rhyme and reason, to comparisons of species of different sizes. The simulation shows bivariate comparisons of tooth length and width for species ranging in size from mouse to elephant. Twenty-six species, A-Z, fill the range with little overlap. Notice how difficult it is to compare raw measurements for species A and Z, and how easy this comparison is when measurements are logged.

Figure 1.

Bivariate plots of simulated tooth length and width for 26 species, A-Z, spanning the range from mouse to elephant. Each species is represented by 50 specimens (gray) with lengths and widths drawn at random from a normal distribution. Model coefficients of variation are 0.05 and the correlation of length and width is 0.3. Each species is represented by the same set of simulated sizes in each plot. Note the confusing change in variance of raw measurements with size, as well as the simple uniformity and comparability provided by transformation to natural logarithms. Here, as elsewhere in nature, functional relationships depend on proportion, not on size.

I focus on size because size is what we measure, and the examples here are univariate for simplicity. However, sometimes two or more variables have to be considered simultaneously as shapes. Similar reasoning applies to combinations in multivariate studies.


Paul Koch[12] made an interesting study of geographic clines in North American mammals, which is relevant to interpretation of spatial variation in the morphology or form of species. Latitude was found to be the best predictor of body, skull, foot, and tail length across a broad range of species, and size change followed Bergmann's Rule. Koch also studied tooth-size clines in five mammalian species. Tooth size was found to be correlated with latitude, cold month temperature, and mean annual temperature.

Koch found that tooth crown area typically changes geographically by 0.20 units, which is about one-half the range of variation in a species at one place and time; tooth crown area can change by as much as the full range. The average rate of change of tooth crown area is about 0.015 natural log units per degree of latitude. The average rate of change of tooth crown area is about 0.017 natural log units per degree of temperature. The amounts of change are substantial, but they are less than those observed in chronoclines.


George Gaylord Simpson introduced the concept of chronoclines in species, paralleling and complementing the older concept of geographic clines.[13, 14] Primate chronoclines have been studied in detail in the early Eocene of the Bighorn Basin[15] and Clarks Fork Basin[16] in Wyoming.

A well-studied chronocline in early Eocene species of Cantius is shown in Figure 2. Two species, Cantius ralstoni and C. trigonodus, were named by Matthew in 1915.[17] With further collecting, three additional species were named preceding these, connecting them, and succeeding them.[18, 19] The five species in Figure 2 span a range of about 10 standard deviation units in size, meaning that successive species pairs differ by about two standard deviation units. Closely related species that differ by this amount are easily distinguished by size alone. Cantius ralstoni and C. trigonodus were named by Matthew, who diagnosed them as having molar tooth rows less than or greater than 14 mm.[17] This would be equivalent to drawing a vertical line in Figure 2 that intersects the horizontal axis at or near zero. Given the record we have now, with many specimens intermediate in size connecting C. ralstoni and C. trigonodus, typological division by an arbitrary measure like this would imply the parallel presence of contemporaneous species through a substantial interval of time — when there was no parallel presence.

Figure 2.

Temporal sequence showing change in tooth size in five species of the early Eocene primate Cantius from the Clarks Fork Basin, Wyoming. Tooth size here is the natural logarithm of first lower molar (M1) crown length multiplied by width (measurements in mm). A total of 459 individual specimens from 38 successive samples are shown, spanning 720 meters of stratigraphic section representing about 1.3 million years and 1.3 million generations of evolutionary time.[16] Cantius ralstoni and C. trigonodus were named by Matthew in 1915.[17] C. mckennai and C. abditus were named by Gingerich and Simons in 1977, when it became clear that C. ralstoni and C. trigonodus were connected by intermediates.[18] C. torresi was added by Gingerich in 1984.19 Note that successive species have mean values differing by about 2 standard deviation units in tooth size. Successive species differ also in the shape of the lower premolars and development of a mesostyle on upper molars (not shown here).

There is a continuum from C. ralstoni to C. trigonodus, but existence of the continuum does not make these species any more the same than they were in Matthew's time. The only way to recognize both differences between successive species and their continuity through time is to draw arbitrary horizontal boundaries. The distinction between species requiring a lineage to be divided is morphological, but the boundary itself is logically chronological. Representation in chronoclines is the most informative way to show phylogenetic relationships graphically.

Species are important because they exist, not as discrete archetypes, but as population pools of genetic and phenetic variation at a place and time: population pools that are recognizable morphologically — by empirical standards — in comparison to other species at that place and time. The better the record of change through time, the more arbitrary is subdivision of lineages into species.

In Fig. 1, species A is not species B nor species C, and so on. Finding intermediates that link these in time will not make them the same. Similarly, in Fig. 2, Cantius ralstoni is not C. trigonodus. Change through time can be substantial, with rates often on the order of 0.1 standard deviation per generation on a per-generation time scale.[20] Evolution is about form, space, and time; it is about variation and change. Interpretation of species in the fossil record touches all of these.

Species are important because they exist, not as discrete archetypes, but as population pools of genetic and phenetic variation at a place and time: population pools that are recognizable — morphologically, by empirical standards — in comparison to other species at that place and time.


  • Philip Gingerich is a professor of vertebrate paleontology at the University of Michigan. His research is focused on interpretation of patterns of variation and rates of evolutionary change in lineages of Eocene primates and other mammals with a dense and continuous fossil record. E-mail: gingeric@umich.edu