Dr. Rob Dillon, Coordinator

Tuesday, September 4, 2012

When worlds collide: Lumpers and splitters

Apparently I have the reputation of being a “lumper.”  The subject came up last month when I was visiting my friend Tim Pearce, curator of mollusks at the Carnegie Museum in Pittsburgh.  Tim asked me a question, which was not actually in the form of a question, but rather phrased as a “philosophy of classification,” upon which he invited me to comment. Tim’s proposition was as follows: 
In the face of uncertainty, it is better to split than to lump.  Because if additional information subsequently becomes available suggesting that two taxa you have split should have been lumped, it will be trivially easy to lump them.  But if additional information becomes available suggesting that two taxa you have lumped should have been split, your entire data set may have been ruined, because information has been lost. 
My first response to Tim was that I’ve never been in such a position.  And my second response was that I could not imagine how such a situation could ever exist.

Is our ignorance complete and our uncertainty absolute, such that we have no existing taxonomy to guide us, no reference works of any sort to fall back on, nor any data to point us in one direction or the other?  And is somebody now holding a pistol to our heads, forcing us to make our decision?  We can’t simply leave the organisms unclassified?

The situation seems to be that we have been transported in shackles to Brazil, and made to classify beetles for our supper, without a guidebook.  I actually could not make my mind address Tim’s premise.

And then it occurred to me that Tim had phrased his proposition not as a question of science, but as a “philosophy of classification."  And it dawned on me that he and I were speaking different languages.

At dinner that evening Tim and I sorted the situation out.  As a museum curator, Tim spends most of his professional life with what we resolved to call a “Worldview of Information.”  He shares this worldview with print publishers, almost everybody involved in the (rapidly becoming omnipresent) internet, and anybody who may be left in the (rapidly becoming obsolescent) library.  Tim’s primary focus is not on the production of information, but on its management, organization, storage, and retrieval.

I spend most of my professional life with a Worldview of Science.  Science is the construction of testable models about the natural world.  The Worldview of Science is not incompatible with that of Information, of course.  When I publish a new hypothesis, or gather data supporting an existing hypothesis, I create information.  But Science is not compatible with an Information Worldview, either.  Because I focus on the quality of a model – old or new, good or bad – with no regard whatsoever for the organization, storage or retrieval of the information that will be generated as a byproduct.

The existence of parallel worldviews, each with its own language, culture, values and assumptions, has been a recurring theme on this blog for quite a few years now.  Most often I have contrasted the Worldview of Science with that of Politics and Public Policy – hit the “Science and Public Policy” tag at right for more [1].  I also posted one essay (back in February, 2011) contrasting Science with the Worldview of Art [2]

Religious faith is another obvious example of a parallel worldview.  I have a great deal of experience in the relationship between Science and Religion, although I have not published on this blog about the subject.  In other fora [3] I have compared that relationship to playing baseball and playing the banjo - neither more valid or more true, neither better nor worse, not incompatible, but not compatible, either.  Simply, profoundly different.

Subatomic particles are too small to see, but physicists can tell they are there when they run into each other.  The various Worldviews I have catalogued above – Art, Science, Politics, Religion and Information – are too big to see.  But one can tell they are there when they run into each other.

I have long derived embarrassing levels of schadenfreude from the Creation/Evolution controversy because I enjoy watching at least three [4] worlds collide - Science, Faith, and Politics.  Sitting in a meeting of the Senate Education Committee hearing testimony on creationist legislation we can watch baseball teams and bluegrass bands trying to drive nails with banjos and catcher's mitts, before a crew of carpenters.  I find this intellectually fascinating.

Science and Religion can collect Information into a three-world collision as well.  A google search on the keyword "evolution," for example, will return a steaming bouillabaisse of Science and Religion mingled in baffling fashion.  My good friend Kelly Smith from the Philosophy Department at Clemson has recently indicted the Information community for aiding and abetting Religion in its ongoing attacks on Science [5].

But the first commandment in Tim’s worldview is this: "Thou Shalt Not Lose Thy Information."  So seen in that light, his question about lumping and splitting put me in the position of a heathen, standing before the Spanish Inquisition.  Who could have expected that?

And my first commandment is this: "Thy Model Shalt be the Best."  As a scientist, I am horrified not simply by the quality of the information on “Wikipedia,” but indeed by the very concept of open publication itself [6].  I've spent a lot of time working with the NCBI GenBank in recent months, for example, and it turns out they will let any bonehead upload any string of the characters A, T, G, and C and call it anything.  The NCBI online resource may be dressed up like Science, but it is no more scientific than the B-Minor Mass.  And I don’t much care for Bach.

Ian Barbour
We'd all probably agree that worldview conflicts, whether they take the form of monkey trials, endangered species panels, or just prickly little interactions between museum curators and evolutionary biologists, are a bad thing.  But are they unavoidable?

My favorite treatment of worldview collision is that of the distinguished philosopher Ian Barbour [7].  Barbour's work specifically addresses the relationship between Science and Religion, but his ideas will generally apply to the relationship between Science and Public Policy, or Science and Information, for that matter.  In addition to conflict, Barbour has proposed three other forms of worldview interaction: independence, dialogue, and integration.

Independence is the model that Steven J. Gould famously called "Nonoverlapping Magisteria [8]."  And independence does indeed describe the actual relationship between baseball teams and bluegrass bands quite accurately.  Nobody has ever tried to bring a banjo into the batter's box, as far as I know.

All of the problems I have outlined above, however, arise from the intellectual appeal of integration.  Preachers do in fact propose science from their pulpits, and scientists do in fact propose public policy in their seminars, metaphorically carrying their banjos into batter’s boxes with the regularity of the tides.  In such situations, the only alternative to conflict is dialogue.

The first steps toward peace between two peoples are taken when those peoples begin to understand that they are, in fact, two peoples.  So I welcomed the dialogue Tim and I had in Pittsburgh last month, and do hope it will continue.  I am more than happy to let colleagues from the Worldview of Information handle all matters of data flow, storage, and retrieval.  We should be pleased if they would leave the Science to scientists.


[1]  Worldview collision is most explicitly addressed here:
Idaho Springsnail Panel Report [23Dec05]
When Pigs Fly in Idaho [30Jan06]
Red Flags, Water Resources, and Physa natricina [12Mar08]
Mobile Basin IV: Goniobasis WTFs [13Nov09]

[2]  When Art and Science Collide [4Feb11]

[3]  Only for those of you who can speak (or at least read) Religion!
  • Dillon, R. T. (2008)  Stonewall, Woodrow, and Me: Reflections on the other great commission.  SciTech, The journal of the Presbyterian Association on Science, Technology, and the Christian Faith 17(3): 7 - 9. [pdf]
  • Dillon, R. T. (2011) Charles Darwin and Theodicy.  (A celebrity death match between Rob Dillon and Francisco Ayala!)  SciTech, The journal of the Presbyterian Association on Science, Technology, and the Christian Faith 20(1): 1-3.  [pdf]
  • Dillon, R. T. (2012)  Science and the Christian Religion: A Sermon in Three Acts.  Preached at Circular Congregational Church, Charleston, SC.  February 12, 2012. [pdf]
[4] I see three worlds colliding because I can understand their three languages.  I am fluent in both Science and Religion, and speak some Politics.  The Worldview of Education is also always represented in the constituency at hearings on creationist legislation, although silently.  On rare occasions somebody seems to attempt a pidgin form of Business and Commerce (“A rigorous science curriculum is necessary to advance South Carolina’s competitiveness.”)  But I don't speak any Money at all.

[5] Smith, Kelly C. (2012)  I Also Survived a Debate with a Creationist (with Reflections on the Perils of Democratic Information).  Reports of the National Center for Science Education 32(2): 6.  [pdf]

[6] The irony of publishing this statement in a blog post does not escape me.

[7] Barbour got his Ph.D. in Physics from Chicago in 1950, swapping over for a B.Div.
from the Yale Divinity School in 1956.  He taught for many years at Carleton College, retiring in 1986.  In 1999 he was awarded a Templeton Prize, punching his ticket on the crazy train.  Barbour has written at least 6 – 8 books in his long career, my favorite being When Science Meets Religion (2000).

[8] Gould, S. J. (1999) Rocks of Ages: Science and religion in the fullness of life.  Ballentine.


  1. "And is somebody now holding a pistol to our heads, forcing us to make our decision? We can’t simply leave the organisms unclassified?" No you can do what a lot of geneticists do. When it comes to 4th and 1 they punt. Seems kind of invertebrate-like to me.

    "But Science is not compatible with an Information Worldview, either. Because I focus on the quality of a model – old or new, good or bad – with no regard whatsoever for the organization, storage or retrieval of the information that will be generated as a byproduct." I would say this is complete nonsense. Systematics AND taxonomy are science. The naming of organisms (wheter it's lumping and splitting) ARE hypotheses that are testable. They also function as a means of communicating science. Where would you be if in the process of gathering data (names) for your FGNA project everyone had punted?

  2. OK, I'll bite.
    If Rob's first commandment is: "Thy Model Shalt be the Best"
    The Best what?
    The Best for what?
    And does Science permit only one answer to this question?
    I'll go out on a limb and posit that different species concepts persit because each "best" serves the needs of a particular community.

  3. Typically the scientific model that yields the most accurate predictions is considered "best." So for example, the Weatherman on our local Charleston Channel 2 uses six models to forecast the track of hurricanes. (Google "spaghetti models" for more!) I feel sure that eventually, as data accumulate matching the actual paths of hurricanes to all the multi-colored predictions available, the strands of spaghetti will converge on a single best model.

    The first scientific model for the origin of species was Charles Darwin's, of course, although a great deal of progress has been made in 150 years, including the development of a firm definition of the word, "species." It turns out that research in this area is hard. So hard, in fact, that a bunch of gene-tree jocks have begun a campaign to change the definition of the word "species" to something easier. It is as though my Channel 2 Weatherman decided to define Carnival Cruise Ships as hurricanes, so that he could predict hurricanes with greater accuracy.

  4. Well put, I agree! Science should be favor models with predictive accuracy. It was your strong stand on lumping vs. splitting that brought to mind cases like this:

    Suppose you have genus with five species. Four seem to have slightly more in common with each other than they do with the fifth species, so a researcher puts the fifth species in its own subgenus. Another researcher then comes along and with the same data elevates the subgenus into a new sister genus.

    What possible empirical test can say who's right? Subgenus or sister genus? To lump or to split?

    If the choice is independent of the data is it inconsequential? Perhaps, and yet the re-printing of every field guide on the subject will hang on the answer.

    So, the "best" nomenclature choice can't simply be about about making predictions and being best at fitting the data. I'm suggesting that when multiple choices can "fit" the data equally well our selection is legitimately about the best communication of the data. That brings us into the realm of shared conventions of usage (which may differ from one social group to another, and which may be unreflectively traditional or deliberately chosen) and what aspects and uses of the data are valued most highly by the group.

    In other words, nomenclature is as much an art as a science. (I was about to say "Science is as much an art as a science", but that seemed like an unecessarily paradoxical way to communicate my point)