Primates are known for their elaborated brains and high encephalization levels, i.e., larger brains than expected for their body sizes (Marino, 1998). The only other mammalian group that rivals primates in this regard is cetaceans (dolphins, porpoises, and whales). Modern cetaceans have been evolving separately from their closest living sister taxa for at least 52 Ma (Gingerich and Uhen, 1998) and from Primates for up to 92 Ma (Kumar and Blair Hedges, 1998). The limited data prior to this study shows that the earliest cetaceans possessed low encephalization levels well below one (Jerison, 1973; Gingerich, 1998; Marino et al., 2000), as was typical of Cretaceous mammals (Jerison, 1973), among which are likely the common ancestor of Primates and Cetacea. Many modern toothed whale (suborder: Odontoceti) species, on the other hand, possess extremely high encephalization levels second only to that of modern humans (Marino, 1998) and therefore have undergone substantial increases in encephalization during their evolutionary history. Likewise, there is evidence for convergent behavioral abilities between odontocetes and humans plus great apes as well (Marino, 2002). These include mirror self-recognition (Reiss and Marino, 2001), comprehension of artificial symbol-based communication systems and abstract concepts (Herman, 2002), and the learning and intergenerational transmission of behaviors that have been described as cultural (Rendall and Whitehead, 2001).
Despite these commonalities, the odontocete evolutionary pathway has proceeded under a very different selective regime from that of primates. Therefore, the highly expanded brain size and behavioral abilities of odontocetes are, in a sense, convergently shared with humans. A description of the pattern of encephalization in toothed whales has enormous potential to yield new insights into odontocete evolution, whether there are shared features with hominoid brain evolution, and more generally how large brains evolve. Yet, up until now, little has been known about the basic pattern of encephalization that characterized odontocete evolution since their adoption of an aquatic lifestyle. The present study provides the first comprehensive description and statistical tests of the pattern of change in encephalization level in cetaceans over 47 million years.
MATERIALS AND METHODS
Scanning and Measurement of Endocranial Volume
Fossil cetacean endocranial volumes were measured from computed tomography (CT) scans acquired with a Siemens Somatom SP scanner at the National Museum of Natural History (USNM), Smithsonian Institution. Image acquisition, analysis, and file conversions were controlled by Siemens SOMARIS software. OSIRIS software was used to convert Siemens image files into DICOM images. Additional scans were obtained at the Medical University of Charleston in Charleston, South Carolina, on a Marconi MX8000 multiple-slice spiral scanner and at Methodist Hospital in Arcadia, California, on a Picker PQ 5000 single-slice spiral scanner. Contiguous 1–2 mm coronal scans of the entire cranium of each specimen were obtained using different scanning parameters depending on the estimated density of the fossil and endocranial matrix, level of permineralization of the bone, and whether the skull was embedded in hardened matrix (Marino et al., 2003).
Each coronal slice through the endocranial cavity was traced manually using Scion Image or ImageJ image analysis software to calculate the area of endocranial space in each slice. The slice areas were then multiplied by the slice thickness and summed to determine the endocranial volume for each specimen, which was used as an estimate of brain mass. Endocranial volume can be converted to mass using the density of brain tissue, approximately 1.0 g/ml, and can be measured directly in modern cetaceans. Previous within-rater and between-rater intraclass correlation coefficients for a subset of archaeocete specimens ranged from 0.91 to 0.98, indicating extremely high agreement across ratings of these specimens. Figure 1A displays an example of a fossil odontocete cranium from the sample and Figure 1B displays an example of a coronal CT image through a fossil odontocete cranium.
The volume of the endocranial space is an overestimate of brain size because cetaceans have an endocranial vascular structure known as the rete mirabile that also occupies this space. From fossil and recent specimens in which the rete mirabile volume can be measured or accurately estimated, it has been shown that the relative size of the rete mirabile has not systematically changed over time (Marino, 1995; Marino et al., 2000).
A total of 66 fossil cetacean crania were scanned and measured. This subset was added to brain and body weight data from a total of 144 modern cetacean specimens for a total sample in the present study of 210 specimens representing 37 families and 62 species. The data for each modern and fossil species in the present study are displayed in Table 1.
Table 1. Data plotted in Figure 1. Rows with geomean listed in the Specimen column are geometric mean values for all of the specimens from single species in a time plane
Superfamily or Infraorder
Brain vol (cc)
Body mass (g)
Species listed with an asterisk (*) were differentiated in an unpublished dissertation. (A. C. Myrick, Jr., University of California, Los Angeles (1979).)
Museum acronyms: CAS, California Academy of Sciences, San Francisco, CA; ChM, Charleston Museum, Charleston, SC; GSP-UM, Geological Survey of Pakistan, University of Michigan collection; USNM, United States National Museum of Natural History, Washington D. C.; YPM, Yale Peabody Museum, New Haven, CT.
Since very few fossil cetacean specimens include entire skeletons from which one can obtain a skeletal length or the multiple anatomical measurements needed to estimate body mass from postcranial elements, we selected the occipital condyle breadth (OCB) to predict body size of odontocetes. OCB was measured on a wide range of modern cetacean specimens with known body masses and was strongly correlated with body mass (r2 = 0.79). This allowed us to use the regression parameters from that analysis to estimate body mass for fossil cetacean specimens when only the cranium was available.
Calculation of Encephalization Quotients
Encephalization is typically expressed as an encephalization quotient (EQ). EQ is an index that quantifies how much larger or smaller a given animal's brain is relative to the expected brain size for an animal at that body size (Jerison, 1973). Brains with EQs larger than 1 are larger than the expected size, while those less than 1 are smaller than the expected size. EQ values were calculated for each specimen in the present study using measured endocranial volumes (or fresh brain weight in some Recent specimens) and estimated body sizes in fossils and actual body weights for all Recent specimens (Marino et al., 2003). The equation EQ = brain weight/0.12 (body weight)0.67 from Jerison (1973) was used to derive EQ values (hereafter referred to as EQ0.67) for each genus or, when possible, each species represented in our sample.
Standard EQ values, like most ratios, are not normally distributed. To avoid the problematic statistical properties of ratios and to be able to perform parametric statistical tests on EQ values, log10EQ0.67 values were calculated. All statistical tests were performed on the logged values, which has the same outcome as taking residuals between logged actual and logged expected brain mass values. EQ values can be calculated using alternative methods to that of Jerison (1973). One way is to derive the regression parameters empirically from the actual sample. The resulting regression equation is the following: EQ = brain weight/1.6 (body weight)0.53. Another popular format for EQ is based on the work of Armstrong (1985), Eisenberg (1981), and others. This approach results in a regression equation of EQ = brain weight/0.055 (body weight)0.75. We calculated EQ for our sample based on these two alternative methods, hereafter known as EQ0.53 and EQ0.75.
Tests of Mean Differences
Each test for a mean difference was conducted using a bootstrap method in which specimens from the groups to be compared were pooled and repeatedly sampled 10,000 times with replacement to produce a distribution of differences between the means. In most comparisons conducted, the actual difference between the two means was outside the range of the bootstrapped distribution of differences.
Tests for Directional Tendencies
A tendency toward increase or decrease in EQ values was tested by reconstructing ancestral states at nodes using parsimony (Maddison and Maddison, 1992), fossil first occurrences (Alroy, 1998), and maximum likelihood (Pagel, 1997) methods using published phylogenies for the Delphinoidea (de Muizon, 1988), Odontoceti (de Muizon, 1991), and Cetacea (Geisler and Sanders, 2003) based on fossil and extant morphology and a phylogeny based on molecular data for Cetacea (Nikaido et al., 2001). For parsimony and fossil first occurrences, nodal states were reconstructed and then compared to the values for adjacent nodes to produce counts of increase and decrease. In the fossil first-occurrence method, each node in the tree was assigned the same state as the descendent taxon with the earliest first occurrence. For maximum likelihood, a likelihood value was calculated for a model of evolution that includes a tendency toward increase or decrease. Then, another likelihood value was calculated for a null model in which no tendency toward increase or decrease exists. Finally, the likelihood value for the model representing a tendency was compared to the likelihood value for the null model to determine if the first is significantly higher than the second.
Figure 2 shows the mean values of EQ0.67 and log10EQ0.67 for the modern and fossil genera or species in the present sample. The range of values measured for modern odontocetes is shown at the top of Figure 2, at 0 Ma. Modern odontocetes have EQ0.67 values ranging from slightly less than 1 to slightly greater than 5, indicating that most modern odontocetes are more encephalized than average mammals at all body sizes occupied by odontocetes. Humans have an EQ0.67 of about 7 on this scale (Marino, 1998).
The most striking finding is that Oligocene odontocetes (mean log10EQ0.67 = 0.312) are significantly more highly encephalized (bootstrap test of difference between means, P < 0.0001) than the Eocene archaeocetes (mean log10EQ0.67 = −0.339), from which they are thought to have been derived (Uhen and Gingerich, 2001). There is no overlap in the range of EQ values between the two groups. The increase in relative brain size from archaeocetes to Oligocene odontocetes reflects both an increase in mean absolute brain size, from 749 (n = 5) to 782 g (n = 5), as well as a decrease in absolute body size, from 1,654 (n = 5) to 207 kg (n = 5).
The mean EQ value for all odontocetes did not change significantly between the late Oligocene (27 Ma, mean log10EQ0.67 = 0.312) and the middle Miocene (14 Ma, mean log10EQ0.67 = 0.404), nor between the middle Miocene and the Recent (0 Ma, mean log10EQ0.67 = 0.402). Figure 2 also shows that by the middle Miocene, almost the full range of modern odontocete EQ values had been achieved. The upper bound in both time periods is formed exclusively by members of the Delphinoidea; in other words, the highest EQ values in odontocetes were only achieved within this specific superfamily of highly derived odontocetes. Middle Miocene delphinoids have a mean EQ (log10EQ0.67 = 0.546) that is significantly higher (P < 0.02) than that of the Oligocene odontocetes (log10EQ0.67 = 0.312), from which they are thought to have been derived. From the Miocene to the Recent (log10EQ0.67 = 0.516), the mean EQ value for delphinoids does not change significantly, although the range increases slightly. Figures 3 and 4 show that the pattern of results yielded by the use of EQ0.53 and EQ0.75 was indistinguishable from data based on the formula described by Jerison (1973).
The fact that significant increases in EQ occurred in odontocete evolution does not necessarily imply the existence of a general tendency toward an increase over the entire clade. A general tendency is a clade-level property, implying in this case a predominance of increases in EQ over decreases among lineages, which would be the expectation, for example, if natural selection consistently favored increases in EQ. In contrast, the absence of a general tendency, i.e., a roughly equal number of increases and decreases, might be the result of selection acting in different directions in different lineages, with no overall clade-level regularity, that is, no net tendency for selection to favor high EQ (McShea, 1994). The tests did not show any significant general tendency toward increase or decrease in any of the phylogenies explored. Increases were slightly more prevalent, but this tendency was not statistically significant. Nor was there any significant difference in the magnitudes of increases and decreases, on average.
The findings of the present study afford the first opportunity to begin to address long-standing hypotheses regarding brain evolution in toothed whales. The results show that encephalization increased in two critical phases in the evolution of Odontoceti. First, the origin of odontocetes from archaeocetes near the Eocene-Oligocene transition occurred contemporaneously with a significant increase in encephalization. Furthermore, because there was no significant increase in brain size over the course of archaeocete evolution, these findings rule out the hypothetical possibility that large relative brain size was associated with the invasion of aquatic habitats.
Another major hypothesis regarding the high encephalization in toothed whales focuses on the neural processing needs associated with either echolocation per se or its elaboration into a complex perceptual system in the suborder Odontoceti (Jerison, 1986; Ridgway, 1986; Oelschläger, 1990). Results here show an increase in encephalization at the origin of Odontoceti that may be related to the emergence and elaboration of the ability to process high-frequency acoustic information associated with echolocation. Echolocation is an ability found in all modern odontocetes, thought to have existed in all known fossil odontocetes (Fleischer, 1976; Fordyce and de Muizon, 2001) and to have been absent in all archaeocetes (Uhen, 2004). Further investigations of echolocatory abilities are currently being undertaken in odontocetes and mysticetes to explore the relationship between changes in encephalization and perception of high-frequency sound as indicated by study of the fine scale anatomy of the internal periotic.
The post-Oligocene period is characterized by little change in the mean encephalization level for Odontoceti as a whole. The origin of Delphinoidea, however, is associated with a significant increase in encephalization over other odontocetes. From the middle Miocene to the Recent, delphinoids form the upper range of encephalization levels and are the exclusive occupants of the upper third of the range of encephalization levels in the Recent.
The present findings provide critical data for further investigations of those factors that may have played a role in the increase in encephalization in delphinoids above the encephalization levels achieved by odontocetes in general. Hypothesized causes of increased encephalization in odontocetes include such varied and not altogether independent factors as social ecology (Connor et al., 1998) and communication (Jerison, 1986). Now that the basic pattern of encephalization change for Odontoceti as a whole has been documented, this pattern can be mapped onto an accepted phylogeny and these other factors can be explored as potential causative factors for the documented changes in encephalization.
The conventional wisdom holds that increased encephalization confers a selective advantage and that increases in encephalization should be pervasive across groups and their component lineages (Gould, 1988). The shift to higher encephalization levels within Odontoceti at the origin of Delphinoidea, however, and the continued expansion (toward both higher and lower levels) within Delphinoidea suggests the absence of an overall drive toward higher levels of encephalization for Delphinoidea as a whole. This does not preclude the possibility that there may have been selective forces acting on individual lineages within the Odontoceti. If increasing encephalization was pervasively advantageous across lineages, however, our tests did not detect it in the available historical record of this group.
The observation that there is a single remaining human lineage that has been pruned down from a bushier tree has led to a popular view that several species of highly encephalized animal cannot coexist spatially or temporally (Tattersall, 2000). Our results show that not only do multiple highly encephalized delphinoids coexist in similar and overlapping environments today, but this situation arose at least as early as the middle Miocene and has persisted for at least 15 million years.
Specimens and CT scanning facilities were provided by the National Museum of Natural History and D. Bohaska, B. Frolich, J. Mead, C. Potter, F. Whitmore, R. Purdy. Access to additional specimens were provided by A. Sanders, Charleston Museum; P. Gingerich, University of Michigan Museum of Paleontology; L. Barnes and H. Thomas, Natural History Museum of Los Angeles County. Additional CT scanning was provided by the Medical University of South Carolina (Charleston, SC) and Methodist Hospital (Arcadia, CA). Emory University students E. Garafalo, N. Pyenson, S. Rotenberg, and B. Shamsai assisted in endocranial measurements. Special thanks to P.M. Novack-Gottshall for ancestral-state reconstructions and maximum-likelihood analyses and E.P. Venit for assistance with the statistical analyses. The authors also thank Todd Preuss for his comments on an earlier draft of the manuscript and Louis Lefebvre for his assistance and support. Funding was provided by the National Science Foundation (to L.M. and M.D.U.) and the Center for the Study of Life in the Universe, SETI Institute (to L.M. and D.W.M.).