Putting Pictures to Data: The Case of the Hypothetical Ancestral Mammal

Guest blogger Richard Javier Stephenson writes...

In 2013, Leary, et al. used genetic and fossil records to describe the history of placental mammals. The research concluded that the common ancestor of all placental orders of mammals originated shortly after the K-Pg extinction event and its progeny diversified rapidly to eventually conquer the world we see now.

However, this reconstruction of the mammalian family tree was not really the focus of popular science media reporting. Instead, the usual headline was about the animal seen in this illustration:

Orignal art by Carl Buell

Orignal art by Carl Buell

This is not a real creature. It is a composite created using over 4000 different phenotypic features and twenty-seven genetic nucleotides between eighty-six fossil and extant species, using analytic software to determine the most likely ancestral forms of these features. The final product included details such as cranial and dental structures, skeletal structure, uterus shape, and even possible sperm shape. It is a reconstruction of an organism based on data of its descendants, rather than the typical paleontological reconstruction based on features of its fossilized body parts. The illustration was created based upon this reconstructed data.

What’s fascinating about such reconstructions is they are not just topics of artistic whimsy.  They’re often subjects of a great deal of scientific investigation and hypothesis-building. They are often used to illustrate evolutionary biology concepts as well, such as contextualizing the evolutionary history of taxa or to illustrate views about the common ancestry of a diverse group. Questions arise, however, in the main goals of these illustrations. What do the reconstructions they are based upon serve for advancing hypotheses in paleontology?

The hypothetical placental mammal and other reconstructed ancestors show two things about depictions of scientific hypotheses. The first of these is in what ways artistic representations depict scientific hypotheses. The second aspect that I often find curious is what the goal of theorizing might be.

How to create fictions

The first puzzle of reconstructions is to what they are “anchored” to. There is some similarity to the reconstruction of a common ancestor and the depiction of some known fossil animal.  What a tyrannosaur or giant sloth looked like in life is something which we will never be known short of time travel or some radical Jurassic Park-style cloning. Scientists and artists must find other ways to justify the ways in which organisms are constructed and depicted in scientific diagrams, museum exhibits, or popular news illustrations.

A key idea to be aware of in the reconstruction of any animal is that of the relationship between homologous and analogous traits. Homologous traits are anatomical features of an organism that are “the same” in origin or structure, even if they may differ in their functions. Analogous traits are those that are similar in their functions, but are not necessarily related anatomically or historically. Vertebrate forelimbs are an often-cited instance of the difference between homologues and analogues traits. The wing of a bat, flipper of a cetacean or hand of a human all serve very different roles in the day-to-day lives of those animals, but they are nonetheless homologues due to their shared form and likely evolutionary common ancestor. By contrast, the wings of bats, birds and insects all facilitate flight, but they have distinct evolutionary histories and underlying architecture. Reconstruction of fossils and hypothetical ancestors depends on understanding of both ideas to create their results. Homology is used to “fill in the blanks” by seeing where known portions of the body relate with one-another. Analogy allows those interested in reconstruction in figuring out what the derived structures might do.

A good example of this relation put to work with extinct animals would be pterosaurs. Their closest living relatives are birds and crocodilians. We can use what we know of those taxa to help understand the anatomy, potential function, and evolutionary history of pterosaurs. To understand how they lived their lives and flew, however, we turn to analogous structures and behaviors found in birds or in bats. This could include how teeth are shaped in other vertebrates in relation to diet, what a flying animal of a certain size might have been hunting in an environment, or how they may have kept themselves airborne on air currents. In these cases, reconstructions of the animal require looking at what other animals have a similar functional application of a form and how that function was applied to its environment.

In the case of reconstructed hypothetical ancestors, the relation between homology and analogy is complicated by context. Since the forms generated by statistical analysis can be a result of correlation between features, the functional history of those features in the descendent groups might not be taken into account. In our mammal example, what is shown is the most likely form based on the fossil record and the extant groups, but not how it lived or how a trait might be adapted to some environment. We can statistically analyze the likely structure of the common ancestor of the bat, dolphin or human forelimbs, but unlike a fossil animal--which has a physical forelimb which we can compare to other forms--we’re left with a statistically generated ‘fossil” whose trait functions we must guess. Such reconstructions often have to depend on known or assumed primitive organisms or fossils, such as many of the shrew-like animals found in the estimated time period of the reconstructed ancestor, potentially hiding any anatomical insights if the emphasis is to make it look more like these likely fossil relatives.

Both sorts of reconstructions require a fair amount of “filling in the blanks” with what is known of homologous traits between animals and analogies in the function of those traits. For artists, this leads to an extra layer of fiction: things like color, behavior, or the way muscles or skin might be layered on skeletal features. These elements have historically even been a good way to see what some of the theories or views held most prominent of the time of a depiction’s recreation. An example of this is swamp-dwelling, tail dragging dinosaurs in the early 20th Century compared to more feathered and active dinosaurs of the early 21st Century. Similarly, older, more lizard-like reconstructions of hypothetical common ancestors for pterosaurs can be compared with more dinosaur-like ones considering recent theories as their status as archosaurs.

This insight in how theory affects reconstruction brings us to the second curiosity of hypothetical reconstructions. What is the justification for such reconstructions?

What justifies the fiction?

One way to view a reconstructed ancestor is as a kind of diagram. It is a way to visualize known data and theories into a form that is easily visible and describable. Much like how a climate study will use the squiggles on a page to create a way of quickly and succinctly understanding a data set, a reconstructed ancestor ideally would do the same for data of common origins. The reconstruction must be something that seems basal enough to have been over time modified to the myriad later forms seen later in the history of life. The problem in this is what it does in the service of the data it is trying to show. The lines on a data figure show the results of an investigation--a way to see the numbers in a simple fashion—unlike a speculative reconstruction.

One problem with this idea is that the reconstructed ancestor is used less as a way to visualize the results and more as evidence for the results, as in many of the article mentioned above. The reconstructed mammal exists in part to show the theory’s results. Whether this is an accurate reconstruction or not does not appear to have much weight in whether contrary evidence (mostly in the genetic record) of pre-K-Pg placental mammals. As Lindberg and Ghiselin noted in their criticism of the hypothetical archetypal mollusc (2003), it is confusing to think that something which is a result of compiling data, comparison, and intuition becomes used as proof or validation of the data itself. That it looks like what it does doesn’t prove the data is what it is.  It simply visualizes the data in a way that hopefully fits other theories or puts them to a test.

However, using the reconstruction simply as a diagram of the results of the paper creates other issues. One is whether the reconstruction even does help visualize data. Complicated “diagrams” often lead to opaque interpretations, making it hard to clearly see whether a hypothesis is supported or not. Reconstructions compound this when comprised of dozens or (in the hypothetical placental mammal’s case) thousands of traits from dozens of other species. While this creature does look like how many of us were likely taught the early ancestor of all mammals after the K-Pg event (small, shrew-like, fuzzy), complications exist due to the complexity of the project. If thousands of traits are being compared, then no single animal form is going to adequately show this in a concise visual manner. Additionally, elements are still being filled-in by the artists who have compiled and created the final illustrations, further obfuscating its use as a data figure.

A related issue is the fact that these piecing together these traits must produce a reasonably likely animal. Such conflicts have arisen in the discussion of segmented animals, how to encapsulate the diverse forms of molluscs in the hypothetical archetypical mollusc (HAM), or how complex something like the last universal common ancestor’s (LUCA) DNA must have been. The need for parsimony in the descriptions of these ancestors is a significant cause of this. The more frequently a trait shows up in separate taxa, the more likelihood it has to being basal. But as more basal features are added, it becomes harder to reconcile them in the final form. This becomes especially problematic when it becomes unclear if a trait is that primitive or easily evolvable.

What we have here is a pair of tradeoffs in reconstructions. The first is between explanation and prediction of data.  The second is between parsimony and comprehensiveness. By finding the right balance, a reconstruction can be a stepping off point for further research. This reconstruction looks like known fossils, but is difficult to fit with genetic evidence placing some placental traits almost forty million years earlier. Resolving this problem should be fruitful for future research.

Apart from areas like this, it is not clear this is what the scientists had in mind with their reconstruction.  And I am not sure what the ultimate way to test these new hypotheses might be, save for attempting to find more fossils which we may not know are actually out there.

How fiction communicates fact

Something we see often in popular scientific literature is artistic reproduction of research discoveries.  This includes fancy three-dimensional diagrams of new molecules, the potential orbital view of some exoplanet, or illustrations of what an archeological site might have looked like in its heyday. What’s fascinating about such reconstructions is they are not just topics of artistic whimsy. They’re often subjects of a great deal of scientific investigation and hypothesis-building. They are often used to illustrate evolutionary biology concepts as well, such as contextualizing the evolutionary history of taxa or to illustrate views about the common ancestry of a diverse group.

Articles about exoplanets typically have only a few points of data to work with, such as mass and orbit, the rest being extrapolation from planetary physics and chemistry.  Source: SO/José Francisco (josefrancisco.org).

Articles about exoplanets typically have only a few points of data to work with, such as mass and orbit, the rest being extrapolation from planetary physics and chemistry.  Source: SO/José Francisco (josefrancisco.org).

At the same time, I worry with the presentation of reconstructions such as the hypothetical ancestral mammal in the popular media articles. They seem to highlight the reconstruction of a hypothetical ancestor as the main interest of the original paper, often not discussing or showing much work done with the reconstruction of phylogeny that it also does. These articles also tend to treat the hypothetical ancestor as a real animal or as something that is basically what the real animal looked like. This is not unlike how in some astronomy articles of exoplanets, artist renditions of a planet are taken to be the “for fact” of what those worlds look like, confusing both the goals of the release and how much is truly known by researchers. These unclear goals of what role the reconstruction serves in the end distract from details of the paper to the general public.

In the end, reconstructions still deserve much of the attention they receive. I like that there is still work to be done in describing the mammalian family tree, and that attempts are being made to do so with a broad range of techniques and sources of data. The phylogeny and reconstruction are interesting in how they utilize fossils, extant animals, genetics, and statistical analysis to generate their results, although some of the results might still have some ways to go in order to account for some of the temporal gaps in the record. At the same time, it probably would have done it some favors to have been a bit clearer on what was the intent of showing a hypothetical primitive form and what role the reconstruction had in illustrating this diversity.


  1. Lindberg, D. R., & Ghiselin, M. T. (2003). Fact, theory and tradition in the study of molluscan origins. Proceedings of the California Academy of Sciences54(22/27), 663-686.
  2. O'Leary, M. A., Bloch, J. I., Flynn, J. J., Gaudin, T. J., Giallombardo, A., Giannini, N. P., ... & Ni, X. (2013). The placental mammal ancestor and the post–K-Pg radiation of placentals. Science339(6120), 662-667.


Richard Javier Stephenson was an animation artist, then a linguist, and is now a graduate student of Philosophy at the University of Cincinnati. His main areas of interest are the philosophy of biology, specifically with regards to systematics, paleontology and evolution; as well as the general practice of science and the role of interdisiciplinary research in science.