All 4 books by Edward Tufte now in
paperback editions, $100 for all 4
Visual Display of Quantitative Information
Envisioning Information
Visual Explanations
Beautiful Evidence
Paper/printing = original clothbound books.
Only available through ET's Graphics Press:
catalog + shopping cart
All 4 clothbound books, autographed by the author $150
catalog + shopping cart
Edward Tufte e-books
Immediate download to any computer:
Visual and Statistical Thinking $2
The Cognitive Style of Powerpoint $2
Seeing Around + Feynman Diagrams $2
Data Analysis for Politics and Policy $2
catalog + shopping cart
Edward Tufte one-day course,
Presenting Data and Information
Arlington VA, June 5, 6
Bethesda MD, June 8
Seattle WA, July 11, 12
Portland OR, July 14
Denver CO, July 17
Minneapolis MN, August 15
Chicago IL, August 17, 18
Click here for more information about ET's course and to register.
Inverted tree of evolution

The NY Times usually has good graphics, but this chart seems inverted. It shows the evolution of chimpanzees, gorillas, and humans from the forest ape of 8 million years ago.

When a "family tree" chart is used, I am used to seeing the passage of time, and successive generations, move down the page. In this chart, it moves upward. Here I found it easier to view when I inverted it, rotating the tree pattern to a root pattern.

After the webpage opens, click the link on the right labeled "Chart: way back when."

[link updated February 2005]

-- David Cerruti (email)


I've checked a dozen-or-so sources this morning and they confirm what I had believed: time charts of this nature always have the present at the top. However, that shouldn't stop us asking "why?" - just as hanging a map of the world upside down changes a number of ingrained perceptions.

I think the evolution analogy is with a tree; what is called "the cone of increasing diversity". There is an interesting discussion on the limitations of the cone in Chapter 1 of Stephen Jay Gould's "Wonderful Life". Although the thesis behind "Wonderful Life' has not stood the test of further debate as well as Gould might have hoped it would, this first chapter is certainly a Tufte- quality dissertation on the graphical misrepresentation of the evolutionary story.

-- Martin Ternouth (email)


Family tree graphics came from genealogy studies. Evolutionary graphics like the one in the New York Times came from evolutionary biologists. This may explain the different convention in orientation. The type of information that they present is the same: they show the flow of genes through time. A family tree is at a smaller scale than most evolutionary graphics. Its units are individuals organisms, whereas evolutionary graphics typically show larger taxonomic units. (Biologists also show even smaller scales with these trees. For example, the complete cell lineage in a developing Caenorhabdtis elegans worm is known.)

The only figure in Darwin's On the Origin of Species is a generic evolutionary tree in the same orientation of the New York Times article. It was probably not the first, but likely one of the earliest and may have been a trendsetter. I would imagine that genealogical trees predate Darwin, but I don't know much about their origins. All types of trees are also displayed horizontally, and in this case they always go left to right, which makes sense given the way we read. The horizontal orientation also makes more typographic sense if you have a lot of taxa on your tree. As for the up or down orientation, you could similarly argue that we read top to bottom. But you could also argue that the tree analogy is so powerful and useful that it is better to go bottom to top. That may have been Darwin's rationale (or instinct) since he makes good use of the tree analogy (see p. 130 in the Origin). To me, the orientation doesn't matter that much, because the branching out of the tree shows passage of time at a glance.

Remember that evolutionary trees like the one in the New York Times are usually hypotheses based on the best available evidence and not hard fact. They are subject to change. Displays of uncertainty on the trees are demanded for scientific journals, but are very rarely shown in popular renditions.

Readers of this board may be interested to see the Tree of Life Web Project, which does an excellent job of compiling all that we know about large scale taxonomy.

-- Anthony Darrouzet-Nardi (email)


Remember that Darwin (and many of those who were enthusiastic about evolutionary theory) was a geologist. In geology, many if not most timelines are constructed with the most recent at the top, the oldest at the bottom.

This is because the convention begins with the diagramming of the geological column and geological profiles. Young rock is on top, older rock lower down in the sequence. (See Steno's law for why.)

-- Mark Hineline (email)


For more on this topic, see Martin J. S. Rudwick, 1976, "The Emergence of a Visual Language for Geological Science, 1769-1840" in History of Science, 14:149-195.

Martin, who directed my doctoral thesis, put a copy of "The Visual Display of Quantitative Information" in my hands when it first appeared and said: here, you'd better read this.

-- Mark Hineline (email)


I've always admired Martin Rudwick's long paper on geological displays. It should be posted in our classics section; does anyone have a reprint that we can scan?

-- Edward Tufte


I have a (signed) copy. I'll get it into the snail mail to you for scanning, as long as I can get it back.

-- Mark Hineline (email)


Evolutionary biologists are not uniform in their use of directionality in trees: they can go up, down, left-to-right, or right-to-left. There is some divergence among subdisciplines: population geneticists tend more toward the time-goes-down version (as in the Times), while systematists tend more toward the time-goes-up version (as Darwin used). There is no universal convention, however, and two or more styles may appear in a single work.

The earlier contributor is almost certainly correct about the geological roots of Darwin's version. The choice of the time-goes-up form follows naturally from the long established geologists'convention of depicting geological strata in that way, which in turn follows from the nature of the strata as seen in section (as in a roadcut). Darwin is explicit about comparing the time intervals of his diagram to geological strata (Origin of Species, 1859, p. 124), and it thus would have made little sense to adopt a convention contrary to the geologists'. Darwin also used the time-goes-up form in notebook sketches from the 1830s, where he referred to the "coral of life" because the "base of branches dead" (i.e. extinct). Robert O'Hara has written about the history of tree diagrams, and several of his papers are available on his website

http://rjohara.net/cv/publications.html

In particular, see "Trees of history in systematics and philology." 1996. Memorie della SocietC Italiana di Scienze Naturali e del Museo Civico di Storia Naturale di Milano, 27(1): 81—88; "Mapping the space of time: temporal representation in the historical sciences." 1996. Pp. 7—17 in: New Perspectives on the History of Life: Systematic Biology as Historical Narrative (M.T. Ghiselin & G. Pinna, eds.). Memoirs of the California Academy of Sciences, 20; and "Representations of the natural system in the nineteenth century." 1991. Biology and Philosophy, 6(2):255—274. The first two are pdf's, but the third is only html and, unfortunately, lacks the figures, so you need to actually get the paper to appreciate it.

-- Gregory C. Mayer (email)


I once saw a nice presentation by Pat Hanrahan that included many interesting historical examples of how abstract tree data has been drawn by biologists.

Come to think of it, this is an example of a powerpoint presentation that I enjoyed.

-- Martin Wattenberg (email)


Mark Hineline, that's very nice, please send the article (to Graphics Press), and we'll scan, post, and return to you.

Best,

ET

-- Edward Tufte


A new thread examines some of the inferential issues in displays that link nouns. See https://www.edwardtufte.com/bboard/q-and-a-fetch-msg?msg_id=0000yO&topic_id=1&topic=

-- Edward Tufte


The James Martin Institute for Science and Civilization ( University of Oxford ) is hosting its first world forum; the topic is "Tomorrow's People: The challenges of technologies for life extension and enhancement."

Many of the presentations from day 1 were quite impressive. It was insteresting to see how each presenter used a slightly different interpretation of the classic "evolution" graphic and how the choice of model changed the framework for the discussion.

Slightly off topic but perhaps interesting to this group, there was an excellent speaker who refused to use PowerPoint out of principle. He said that as a philosopher he did not use, or need, any visual aids that were not familiar to Plato. It was brilliant lecture with plenty of rhetoric but no mumbo jumbo.

QUESTIONS FOR THE PANELLISTS

http://www.worldforum2006.net/index.php

-- Tchad (email)


Responding to David Cerruti's original question, I have no trouble dealing with trees that read down or up or left-to-right, or geological sequences reading up or down, l-to-r or r-to-l. I've never encountered an evolutionary tree that read right-to-left, but I'm sure I'd be able to deal with that too. More mundanely, my fingers never get confused between the "1-2-3 at the top" keypads of telephones and the "1-2-3 at the bottom" keypads on a computer. Clearly we humans are able to switch modes quite easily. And considering how many bad design decisions are adhered to because "that's the standard", I think worrying about the vernacular direction can take a back seat to other concerns; we each of us have several vernaculars at our disposal.

I've also seen evolutionary trees arranged from a centre outwards: the decendants of a common ancestor fan around in an almost complete circle. A small arc is left blank between the two ends of the list of latest descendants, lest the least closely related be mistaken for neighbours.

-- Derek Cotter (email)


A little learning can overcome a broken convention, but conventions yield useful economies. Standard formats help analysts devote time to learning about the content rather than learning about a new format. Standard formats allow implicit comparisons to previously seen data in that format, and they assist comparative mental imaging.

All the different date formats are fairly easy to interpret, but what a mess they are (as anyone who writes parse tables knows). Similarly with different units of measurement (feet vs. meters). Yes we can understand both, but the value of one fixed universal convention here would be substantial (and would have avoided some disasters as well).

Evidence presentations that break convention may make readers suspicious about the display. How come north is not at the top of the map? Why does evolutionary time flow to the left, not the right? Don't they know better? Nonbeneficial broken conventions invite the reader to argue with the presenter about graphic methods.

Of course presenters might wish to innovate in their displays. But that is a different topic from the content argument being advanced in the presentation. Perhaps unconventional analytical graphics should be tried out on this board first!

The principle probably should be to stick with convention unless there are clear analytical gains that result from a broken convention. Strunk and White provide similar advice with regard to syntax and writing styles.

A part of the circular evolutionary tree is shown in Beautiful Evidence.

-- Edward Tufte


I have come across all the orientations of evolutionary tree that Derek Cotter mentions (including ones that fan out from some point vaguely in the middle. I've also seen (and drawn) ones that read from right to left. This wasn't just for the sake of variety: I wanted to compare the results from what was generally considered a poor method of analysing data with those from a "good" method, and to show that they were almost exactly the same. Rather than just drawing two trees side by side or one over the other, I thought the message came over more clearly if I listed the organisms in a central column and showed the two trees building out from this column. Inevitably this meant that time went from left to right in the left-hand tree, and from right to left in the right-hand tree, but I didn't (and don't) think that this made the trees difficult to understand.

This was more than 20 years ago, and I can't find an example, but I have an example of a somewhat similar case that was in the Journal of Theoretical Biology in 1983 (101, 317-319). There I was concerned with an idea that was very widespread among experts that phenetic methods (methods based on "distances" between pairs of taxa) were bad because they used only part of the information, and I wanted to point out that parsimony methods (which were much more favoured by experts because supposedly they used all the information) also discarded part of the information. To show this, I constructed two trees by a simple phenetic method, one using all the loci in seven ribonuclease sequences, the other using only the loci that did not contribute to the parsimony solution. If the parsimony enthusiasts were right this second tree should have been random gibberish, but although clearly inferior to the other it was far from being random gibberish. To make the comparison I listed the ribonucleases in the middle column and put the trees around them. I've redrawn the result (quickly and not very precisely) and put it at

     http://bip.cnrs-mrs.fr/bip10/images/rnase.png

One of the general points that emerges from this thread is that there is no standard convention for orienting evolutionary trees, and thus none to be broken. Even genealogical trees are sometimes drawn with the root at the bottom (like real trees, of course). For example, in the sort of books that are given to children to write in the names of their parents, grandparents, brothers and sisters the space for the child's own name is quite likely to be at the top of the page rather than at the bottom.

When a convention does exist I do of course agree that one shouldn't break it without having a sensible reason. When I first came to live in Marseilles I was often confused by city plans, because plans of Marseilles never put north at the top, but usually put it at the left. I've never come across a reason why this should be. It may be because the north-south axis is longer than the east-west axis, but that sort of logic doesn't seem to be applied anywhere else: I haven't seen maps of California that put north at the left, for example, let alone in more extremes cases like Chile. You do sometimes see maps of the world printed in Australia with south at the top, but then there is clearly a message being conveyed.

-- Athel Cornish-Bowden (email)


Professor, as a "Brit", I must strongly disagree that date formats are easy to handle; they're quite intractable compared to the simple code-switching tasks I mentioned, as I discover whenever I have to spend time mentally untangling the dates in a table from an American source into our European big-endian scheme. 11/24/2005 is simple enough, but what of 11/04/2005?

I concede that the South American (now ISO) little-endian convention is best of all, and I use it for all my file names.

But however I quibble with your particular example, I do have some sympathy with the general point. It's just that my sympathy has a limit: I think it's for those who have an intolerance of multiple standards to give up their favourite and adopt that of someone more tolerant. David Cerruti's reaction on learning that some folk draw descent charts from the bottom up should have been "Quick, for the sake of standardization, we must abandon our top-down convention and use this one instead!", not "Only one convention among many arbitrary possibilities is acceptable. Therefore I declare that it should be mine; kindly alter yours." The latter seems to want both the cake and the satisfaction of having eaten it.

-- Derek Cotter (email)


I've rather belatedly figured out how to put images in postings. This one illustrates my posting of 16th March:

Two trees

-- Athel Cornish-Bowden (email)


I agree with Derek Cotter about dates. Even 40 years after I first came across the US way of writing dates I still find it takes some mental effort to interpret. When I first started hearing the phrase "9/11" it took a long time before it stopped provoking the thought what's special about the 9th of November?

On the other hand I think he's unjust to David Cerruti. I don't read his comment as a demand that everyone should draw trees the way he likes (with its root in the sky), only a suggestion that he finds it easier if they are drawn like that.

-- Athel Cornish-Bowden (email)


I think you're right, and now I read again I see that David actually did have trouble interpreting the graph. As soon as someone has difficulties with a format, all responses of the form "well, I don't have any difficulty" should be immediately understood as irrelevant. It's not the problems of the precocious we're trying to address when we seek excellence in clear design, or there'd be no reason not to present text upside down.

-- Derek Cotter (email)


Speaking of evolutionary trees, the American Museum of Natural History reproduces Darwin's first tree from his notebooks of 1837. The tree is bottom-up, except for the branch that goes off to the right. Above his diagram he writes, "I think"...

I don't suppose Darwin was the first biologist to render the Linnaean enterprise as a tree: the hierarchy naturally lends itself to such a diagram, although it wasn't devised with evolution in mind.

-- Derek Cotter (email)



Here is a page from George Hunter's 1914 Biology Book,
A Civic Biology - The book from the State v. John Scopes
case.


http://www.law.umkc.edu/faculty/projects/ftrials/scopes/evolut.htm

How different would the issues be now, in 2006.
What would the visuals look like?

-- Tchad (email)


Three interesting diagrams in this article in scienceblogs.com, discussing a paper that seeks to demonstrate the existence of six or more giraffe species, and not one. The diagrams are:

1A) a cladogram that goes right to left, to the left of which are five giraffe hide patterns. To the left of that is a map of Africa, tying the cladogram, the patterns, and the locations of the giraffe populations together.

1B) a minimum-spanning network of haplotypes, combining a sort of Underground Map arrangement of connecting lines with a bubble chart presentation of haplotype frequency (bigger bubbles more frequent in the population)

2) a neighbour-joining network of allele-sharing distances, presented as a tree with no up, down, left, or right direction, but only outward in whatever direction conveniently offers some free space on the page.

-- Derek Cotter (email)




Threads relevant to analytic design:

Seeing Around: New ET essay published