I’m gathering all six parts of last year’s Tutorial 19 in one place for easy reference. Here they are:

Enjoy!

LACM Deinonychus claw

All I want to do in this post is make people aware that there is a difference between these two things, and occasionally that affects those of us who work in natural history.

In one of his books or essays, Stephen Jay Gould made the point that in natural history we are usually not dealing with whether phenomena are possible or not, but rather trying to determine their frequency. If we find that in a particular population of quail most of the birds eat ants but some avoid them, then we know some things: that quail can tolerate eating ants, that quail are not required to eat ants, and that both strategies can persist in a single population.

This idea has obvious repercussions for paleoart, especially when it comes to “long-tail” behaviors. I dealt with that in this post, and also in the comment thread to this one. But that’s not what I want to talk about today.

Sometimes it is useful to talk about things that never happen, or that have at least never occurred in the sample of things we know of. Obviously how certain you can be in these cases depends on the intensity of sampling and the inherent likelihood of a surprising result, which can be hard to judge. If you argued right now that T. rex lacked feathers because no T. rex specimens have been found with feathers, you’d most likely be wrong; it is almost certainly just a matter of time before someone finds direct evidence of feathers in T. rex, given the number of T. rex specimens waiting to be found and the strength of the indirect evidence (e.g., phylogenetic inference, analogy: ornithomimids are known to be feathered even though most specimens are found without feather impressions). If you argue that sauropods are unique among terrestrial animals in having necks more than five meters long, you’re most likely right; being wrong would imply the existence of some as-yet undiscovered land animal of sauropod size, or with seriously wacky proportions (or both), and our sampling of terrestrial vertebrates is good enough to make that extremely unlikely.

LACM baby rex snout

The reason for this post is that sometimes people confuse that last argument, which is about sampling and induction, with the argument from personal incredulity.

For example, in our no-necks-for-sex paper (Taylor et al. 2011), we included this passage:

Sauropoda also had a long evolutionary history, originating about 210 million years ago in the Carnian or Norian Age of the Late Triassic, and persisting until the end-Cretaceous extinction of all non-avian dinosaurs about 65 millions years ago. Thus the ‘necks-for-sex’ hypothesis requires that this clade continued to sexually select for exaggeration of the same organ for nearly 150 million years, a scenario without precedent in tetrapod evolutionary history.

One of the reviewers argued that we couldn’t include that section, because it was just the argument from personal incredulity writ large, like so:

There are no other known cases of X in tetrapod evolutionary history, and therefore we don’t believe that the case in question is the sole exception.

…with the second part of that unstated (by us) but implied. But we disagreed, and argued (successfully) that it was an argument based on sampling, like so:

There are no other known cases of X in tetrapod evolutionary history, and therefore it is unlikely that the case in question is the sole exception.

Now, it is perfectly fair to criticize arguments like that based on the thoroughness of the sampling and the likelihood of exceptions, as discussed above for T. rex feathers. Just don’t mistake arguments like that for arguments from personal incredulity.* On the flip side, if someone makes an argument from personal incredulity, see if the same thing can be restated as an argument about sampling. Maybe they’re correct but just expressing themselves poorly (“I refuse to believe that the moon is made out of cheese”), and maybe they’re wrong and restating things in terms of sampling will help you understand why.

* If you want to get super pedantic about it, they’re both arguments from ignorance. But one of them is at least potentially justifiable by reference to sampling. Absence of evidence is not necessarily evidence of absence, but it may get to be that way as the sampling improves (e.g., there is no evidence of planets closer to the sun than Mercury, and at this point, that is pretty persuasive evidence that no such planets exist).

LACM brachiosaur humerus with Wedels for scale

Parting shot: one thing that has always stuck in my head from Simberloff (1983) is the bit about imagining a large enough universe of possible outcomes. And I’ve always had a perverse fascination with Larry Niven’s “Down in Flames”, in which he pretty much demolished his Known Space universe by assuming that every basic postulate of that universe was false. Neither of these follow directly on from the main point of the post, but they’re not completely unrelated, either. Because I think that they yield a pretty good heuristic for how to do science: imagine what it would take for you to be wrong–imagine a universe in which you are wrong–and then go see if the thing that makes you wrong, whatever it is, can be shown to exist or to work. If not, it doesn’t mean you’re right, but it means you’re maybe less wrong, which, if we get right down to it, is the best that we can hope for.

The photos have nothing to do with the post, they’re just pretty pictures from the LACM to liven things up a little.

References

As I noted in a comment on the previous post, titanosaurs have stupid cervicals.

As evidence, here is as gallery of titanosaur cervicals featured previously on SV-POW!.

1. From Whassup with your segmented lamina, Uberabatitan ribeiroi?, an anterior cervical of that very animal, from Salgado and Carvalho (2008: fig. 5). As well as the titular segmented lamina, note the ridiculous ventral positioning of the cervical rib. It’s like it’s trying to be Apatosaurus, but it just doesn’t have the chops.

uberabatitan-cervicals-copy

2. From Mystery of the missing Malawisaurus vertebra, this alleged vertebra of that taxon from Jacobs et al. (1993:fig. 1), which completely fails to resemble all the other cervicals subsequently described from Malawisaurus (see the earlier post for details). Note the crazy sail-like neural spine and super-fat parapophyseal stump.

malawisaurus-1993

3. From Futalognkosaurus was one big-ass sauropod, this completely insane posterior cervical vertebra of Futalognkosaurus in right anterolateral view, with Juan Porfiri (175 cm) for scale. It’s super-tall — much taller than it is wide, and seemingly taller than it is long.

Posterior cervical vertebra of Futalognkosaurus in right anterolateral view; Juan Porfiri (175 cm) for scale

4. From Ch-ch-ch-changes, cervical 11 of Rapetosaurus, from Curry Rogers (2009:fig. 5). Notice how tiny the centrum is compared with the tall superstructure, and how the neural spine has such a distinct peak. Weird.

Rapetosaurus cervical

5. From Talking about sauropods on The Twenty-First Floor, cervical 9 of the same Rapetosaurus individual, from Curry Rogers (2009:fig. 9). The neural spine is a completely different shape from that of C11, but that is presumably mostly due to damage. One of the interesting things here is the apparent lack of pneumatic foramina in the centrum. They’re there somewhere: Curry Rogers (2009:1054) writes “In cervical vertebrae 9, 11, and 12, the centrum bears an elongate shallow pneumatic fossa with two anterior pneumatic foramina surrounded by sharp, lip-like boundaries.” But they are hard to make out! 

CurryRogers2009-rapetosaurus-fig9-C9

The meta-oddity here is that the cervicals of the four titanosaur genera pictures here are all so different from each other. What does this mean?

Probably only that Titanosauria is a huge, disparate, long-lived clade that encompasses far more morphological variation than (say) Diplodocidae. It’s a truism that we don’t, even now, really have a handle on titanosaur phylogeny — every new study that comes out seems to recover a dramatically different topology — so our perception of the clade is really as a big undifferentiated blob. In contrast, the division of Diplodocoidea into Rebbachisaurids, Dicraeosaurids and Diplodocids (plus some odds and ends) is nicely established and easy to think about.

So. Lots of work to be done on titanosaurs.

References

I was reading Stephen Curry’s excellent summary of Monday’s Royal Society’s conference on “Open access in the UK and what it means for scientific research”. One point that Stephen made is:

[David Willetts's] argument is that pursuance of green OA leads to an unstable situation in which the cancellation of subscriptions (because readers have free access) drains the system of the funds needed to manage peer review and other publishing costs.

As an analysis of the difficulties of Green OA, this is admirably precise. But my eye was caught by that phrase “funds needed to manage peer review and other publishing costs.”

I think we should make an effort to wean ourselves off the habit of talking about “managing peer review and other publishing costs”. We all recognise that publishers do not provide peer-reviewwe do. But it’s also true that publishers don’t manage peer-review, either. Once again, we do that, by acting as unpaid academic editors.

Make-Money-online

I know that this is not news. We all know this. But a habit of speech is affording publishers a degree of credit that their efforts don’t merit, and that clouds the debate. Let’s apportion credit where it belongs.

Of course there are still “other publishing costs”. These are real and not negligible (even though PeerJ’s financial model suggests they are much less than we have sometimes assumed). It’s right that we should acknowledge that there really are publishing costs; and that whatever financial model we end up will need to pay them somehow. But let’s make an effort to be more precise about what those publishing costs are. Managing peer-review is not one of them.

Mathematician David Roberts has pointed me to a useful new five-part series by Martin Paul Eve, entitled Starting an Open Access Journal. It’s well worth a look, for how it engages with so many practicalities and how tractable he makes it all seem.

We’re actually pretty well served for open-access journals in our field (Acta Palaeontologica Polonica, Palaeontologia Electronica, PalArch’s journal of vertebrae palaeontology, soon PaleoBios, and of course PLoS ONE). But for scientists in other fields that have fewer options, starting their own journal may well be the single most effective thing they can do to advance open access. (It’s going to look pretty good on the CV, too!)

I am briefly quoted in Times Higher Education‘s new article about the White House public access petition Since my response had to be quite dramatically cut for space, here is the full text of what I sent the writer, Paul Jump:

The success of this petition is important for several reasons. First, it puts paid to the pernicious lie that open access isn’t important because research is useless to non-specialists. Support for the petition has come from many non-academic quarters, including patient support group Patients Like MeWikipedia, Creative Commons, the American Association of Law Libraries, and the Association of College and Research Libraries.

Perhaps equally important, it’s attracted support from publishers — not only open-access publishers such as PLoS, BMC and InTechWeb, but also forward-thinking subscription publishers like Rockefeller University Press. It’s also featured widely in the non-academic media, appearing on the news-for-nerds sites SlashdotReddit, and Hacker News, in newspapers like the Guardian, and in magazines like Wired.

All of this makes the crucial point that open access isn’t just an esoteric preference of a few disgruntled academics, as the hugely profitable commercial subscription-based academic publishers have consistently tried to paint it. It’s something that has huge implications for all of our lives: for health care, education, legislative deliberations, small businesses, and ultimately the health of the planet.

Open-access advocates have seen this for a long time, but now the message is getting out. Irrespective of what response the Obama administration makes to the petition’s very rapid achievement of the required 25,000 signatures, what’s been said about it around the world lays waste the idea that open access is nothing more than an alternative business model for scholarly publishing. It’s a much bigger revolution than that.

(They managed to cut that down to 69 words!)

Research Councils UK is the aggregate of the UK’s seven research councils, which makes it overall  the most important and influential funding body for science in Britain.  A few days ago, they released a draft of their new open access policy, and they are soliciting comments now.  Comments can be from anyone — individuals or groups, British or overseas — like the recent OSTP Request For Information in the States which we have to assume was influential in the defeat of the RWA.

What does the new draft policy say?  I urge you to read it yourself: its only six pages, and they are very clear.  To  quote from the document’s introduction:

Key differences with the current policy include:

  • Specifically stating that Open Access includes unrestricted use of manual and automated text and data mining tools; and unrestricted reuse of content with proper attribution.
  • Requiring publication in journals that meet Research Council ‘standards’ for Open Access.
  • No support for publisher embargoes of longer than six months from the date of publication.

Unsurprisingly I am very much in favour of the proposed changes to the RCUK policy, and I will be a making a lengthy submission commenting on individual changes.

But any comment is significant — even you’re just writing to say “I approve of the policy changes”, or “I recommend a 12-month embargo period instead of six” or indeed “open access is a silly fad, this policy in unnecessary”.  The point is that RCUK want to hear your opinion.  Your voice matters.

So please email your comments to communications@rcuk.ac.uk with the subject “Open Access Feedback”.  I am told (by Cameron Neylon, who heard it from a colleague who’d had it from a friend of his) that the deadline is Thursday 5 April, but since that is hardly a definitive source, I recommend getting your submissions in as soon as possible.  (You can be sure that the publishers will be doing so.)

Those ostrich necks I went to Oro Grande to get last Thursday? Vanessa and I started dissecting them last Friday. The necks came to us pre-cut into segments with two to three vertebrae per segment. The transverse cuts were made without regard for joints so we got a bunch of cross sections at varying points through the vertebrae. This was fortuitous; we got to see a bunch of cool stuff at the cut faces, and those cut faces gave us convenient avenues for picking up structures and dissecting them out further.

In particular, the pneumatic diverticula in the neck of this ostrich were really prominent and not hard at all to see and to follow. The photo above shows most of the external diverticula; click through for the full-resolution, unlabeled version. The only ones that aren’t shown or labeled are the diverticula around the esophagus and trachea (which had already been stripped off the neck segments, so those diverticula were simply gone), those around carotid arteries, which are probably buried in the gloop toward the bottom of the photo, and the intermuscular diverticula, of which we found a few in parting out the dorsal and lateral neck muscles.

There is one final group of diverticula that are shown in the photo but not labeled: the interosseous diverticula that fill the air spaces inside the bone.

We have tons of cool photos from this dissection, so expect more posts on this stuff in the future.

For previous posts showing diverticula in bird neck dissections, see:

Things to Make and Do, part 7: fun with rhea necks

Things to Make and Do, part 7b: more fun with rhea necks (admittedly, not the most creative title ever)

Denver Diplodocus

November 5, 2011

Taken by me–or rather, my camera in automatic mode–earlier today, because the ole sauropod blog has been a bit light on sauropods lately.

I spend a lot of time thinking about Sauroposeidon, Supersaurus, and the like. It’s good to be reminded that even an ‘average’ sauropod like Diplodocus is still pretty awesome. And weird. I don’t know if we can be reminded often enough.

Preparing a talk is a time-consuming process, and there’s no question that getting the slides ready is where the bulk of that time goes.  But unless you understand exactly what it is that you’re going to talk about, even the best slides won’t rescue your talk from mediocrity, so before you fire up PowerPoint, go and read part 1 of this tutorial, on finding the narrative.  Seriously.  The slides are how you convey your message, and they’re important.  But not as important as what your message is.

Assuming you know what story you’re trying to tell, here is the overriding principle of slide design: make yourself understood.  Remember again that you have something less than twenty minutes in which to make your rich, complex research project understood to a hall full of strangers who have just sat through five or ten or fifteen other presentations.  They will be mentally tired.  Help them out.  Make every slide tell a clear story.

The slides for a conference talk are science, not art.  That doesn’t mean they have to be ugly — of course it doesn’t.  But it does mean that whenever you find yourself facing a choice between clarity and beauty, go with clarity.

That means you do not want your slides to look like this:

OK, that is not even beautiful.  But it does illustrate some horrible mistakes, and we’ll touch on all of them  in what follows.  For now, just remember that the purpose of a Results slide is to help the audience know what your results were.

So how do you make yourself understood?

1. Use the full size of the screen

Most importantly, don’t “frame” your content.  You have a specific amount of space in which to present your work.  Don’t throw any of it away.  Although the super-bad slide above may look extreme, I have seen plenty to slides that present, say, specimen photos in about the same amount of space as the graph above occupies.  So, then:

  • No picturesque borders.
  • We don’t need the talk title, or your name or address on every slide.  You can tell us once at the start of the talk and then, if you like, once more at the end.  If we truly forget who you are in the middle, we can always look at the programme.  If we forget what you’re talking about, then your talk has more profound problems.
  • That goes double for logos.  We do not need to see the following more than once (or indeed once):
    • Your institution’s crest
    • The conference logo
    • Logos of funding bodies

We don’t need any of that stuff, and all of it wastes precious real-estate.  Space that you could be using to tell your story.

Most important of all: use as much space as you can for your images.  Specimen photographs, interpretive drawings, reproduced figures from the literature, graphs, cladograms, strat sections — whatever you’re showing us, let us see it.

In my own talks, I like to make the picture fill the whole slide.  You can usually find a light area to put a dark text on, or vice versa.  I often find it’s useful to give the text a drop-shadow, so that it stands out against both light and dark background.  (You can find that option in Format -> Character… -> Font Effects if you use OpenOffice, and no doubt somewhere similar in PowerPoint.)

If the aspect ratio of an image that I want to use is not the 4:3 that projectors give you, then I will often crop it down to that aspect ratio, if some of the edges of the image are dispensable, so that the cropped version is properly shaped to fill the screen.

(On image resolution: most projectors seem to be 1024 x 768, maybe some these days are 1280 x 960.  There’s no point using images at a higher resolution than that: your audience won’t see the additional information.)

2. Legibility

Hopefully you won’t need too many words on your slides, since you’ll be talking to us about what we can see.  But what words you use, we need to see.  Specifically, this means:

  • Use big fonts.  There is absolutely no point in showing us an eighty-taxon phylogenetic tree: we just won’t be able to read the taxon names.  I tend to make my fonts really big — 32-point and up, which actually is probably bigger than you really need.  But you don’t want to be smaller than 20-point at the absolute minimum.
  • Use high contrast between the text and background.  That usually means black on white, or (if you must) white on black.  Well, OK — it doesn’t literally have to be black, but it needs to be a very dark colour (I often use very dark blue).  And it doesn’t literally have to be white, but it needs to be a very light colour.  (I occasionally use a very pale yellow “parchment”-type colour, but less often.)  Do not use grey text or a grey background.  Especially do not use grey text on a grey background, even if they are fairly different greys and the muted effect looks classy.  You’re not shooting for “classy”, you’re shooting for “legible”.  Because you remember the prime directive that you’re trying to make yourself understood.
  • If for some reason you must use a non-black, non-white text or background, don’t make it a highly saturated colour.  Some combinations, such as a red on blue, and virtually impossible to read.
  • No vertical writing (with the possible exception of short y-axis labels on graphs).  If your cladogram’s taxon names are vertical, turn your cladogram around.  Redraw it if necessary.  If the audience have their heads on one site, you’re doing it wrong.

3. Font Choice

Apart from size, what else matters about fonts?

  • Avoid elaborate fonts, such as the URW Chancery L Medium Italic that I used for my name and affiliation in the Bad Slide at the top.  They’re hard to read, and at best they draw attention away from the message to the medium.
  • Pick a single font and stick with it for consistency.  Or if you wish, one serifed font (for body text) and one sans-serif (for headings).  But you should have little enough text on your slides that it’s practically all headings.
  • Stick to standard fonts which you know will be on the computer that will be displaying your presentation.  In practice, the safest approach is it stick to Microsoft’s “core fonts for the web” — which is plenty enough choice.
  • You might want to avoid Arial, which is widely considered particularly ugly.  Other ubiquitous sans-serif fonts include Trebuchet and Verdana, which are both rather nicer than Arial (though Verdana’s glyphs are too widely spaced to my eye).
  • Do not use MS Comic Sans Serif, or no-one will take anything you say seriously.  I don’t just mean your talk, I mean ever, for the rest of your life.

Why is it important to stick to standard fonts?  Because of size, spacing and positioning.  Your computer may have the super-beautiful Font Of Awesomeness and it might make your slides looks beautiful; but when you run your PowerPoint file on the conference computer, it won’t have Font Of Awesomeness, so it will substitute whatever it thinks is closest — Arial or Times or something.  Not only will you not get the visual effect you wanted, but the glyphs will be different sizes, so that your text will run off the edge of the page, or fall right off the bottom.

(Handy household hint for users of Debian GNU/Linux and variants such as Ubuntu.  Make sure that you have the MS core fonts installed on your computer, so that OpenOffice can properly display your slides as you’re designing them, rather than substituting.  sudo apt-get install ttf-mscorefonts-installer, restart OpenOffice, and you’re good to go.)

4. How many slides?

I need to mention this issue, if only to say that there’s no right answer.  I don’t say that lightly: for most slide-design issues, there is a right answer.  (Example: should you use MS Comic Sans Serif?  Answer: no.)  But number of slides has to vary between people to fit in with presentation styles.

I tend to use a large number of slides and whiz through them very quickly — my SVPCA 2011 talk had 80 slides, and in 2010 I had 92 slides.  Lots of them are parenthetical, sometimes just a silly joke to make in passing a point that I am already making.  If you miss such a slide, it doesn’t really matter: it’s just light relief and reinforcement, not an integral part of the narrative.

.

But that many-slides-slipping-quickly-past style doesn’t suit everybody. In the eighteen minutes or so that you get to give a talk (allowing a minute for messing about getting set up and a minute for questions), getting through 80 slides in those 1080 seconds gives you an average of 13.5 seconds per slide.

Lots of people prefer to use fewer slides and talk about them for longer. You can give an excellent talk with very few slides if that approach comes naturally to you: step slowly through nine slides, talk about each one for two minutes.

Once you’ve given a few talks you’ll know which approach works best for you, and you can design accordingly. For your first talk, you’re probably best off aiming initially somewhere in the middle — thirty or so slides — and then seeing what happens when you dry-run the talk. (We’ll discuss that next time around.)

5. Miscellaneous

I’ve touched on this one already, but it’s best to use as little text as possible. That’s because you want your audience listening to your story, not reading your slides. I used to put a lot of text in my slides, because I wanted the PowerPoint file to stand alone as a sort of a record of the talk. But I don’t do that now, because a talk involves talking (clue’s in the question). I include enough text to remind myself what I want to say about each slide (sometimes just one or two words; often none at all). And I try to make sure there’s enough to let the audience know what they’re looking at if I zoom straight past it. For example:

.

I used this slide to briefly tell a typical taphonomic story of a sauropod neck.  But I didn’t need to say that I was using diagrams of the neck of Sauroposeidon taken from Wedel et al. 2000, so I just shoved that information on the slide for anyone who was interested.  That way I didn’t have to break the flow of my narrative to impart this information.

Use a consistent colour palette.  If you’ve used dark blue text on white for half of your slides, don’t switch to black on pale yellow for the other half.  It’s not a hugely important point, but it all contributes to helping the talk go down smoothly.  You’re getting rid of mental speed-bumps that could stop your audience from giving their full attention to the story you’re telling.

Where possible, avoid putting important information at the bottom — in, say, the lower 10-15% of the slide.  That’s because the lower part of the screen can sometimes be obscured by the heads of the people in the front rows.

Avoid hatching, which can look terrible on a screen, in a way that’s very hard to predict.  In the Sauroposeidon taphonomy slide above, for example, the lost bones are “greyed out” using a flat grey colour rather the close diagonal lines of the original.  I knew it would look right on the screen.

Skip the fancy slide transitions, animated flying arrows, and suchlike. It’s just distracting nonsense that no one in the audience (or anywhere else, for that matter) needs to be exposed to. It’s just gross. Also, as with fonts, you may end up giving your talk from a machine with an older version of PowerPoint that doesn’t support the turning of animated pages and the bouncing arrival of arrows and clipart, and then your presentation will either look stupid or fail to run entirely.

You might want to draw highlighting marks on your slides, e.g. circles around the relevant parts of a specimen photos.  That will save you having to mess about with the laser pointer later.  (I will have much to say about the laser pointer in part 4).  I like to show two consecutive slides: one of the unadorned photo, then one that’s identical apart from the addition of the highlight, like this:

Then as I am talking about the first slide, “in order to mount the vertebrae in something approaching a straight line, they had to leave a huge gap between consecutive centra”, I’ll step on to the next one, which highlights what I’m saying.  Slick, no?  (This is part of why I end up with such high slide counts.)

A pet hate: don’t write “monophyletic clade”.  If it’s a clade, it’s monophyletic by definition.  “Monophyletic clade” is like “round circle”, “square square” or “boring ornithopod”.

And finally …

Show us specimens.  We are vertebrate palaeontologists, and we love vertebrate fossils.  No-one goes into the field because of a deep and abiding passion for graphs or for tables of numbers.  We understand that from time to time you’ll need to show us those things in order to tell the story, but nothing makes an audience happier than big, clean photos of beautiful specimens.

Well, that’s it — how to make good slides.  Next time we’ll look at rehearsing the talk.  (It’ll be a much shorter post than this one.)

Follow

Get every new post delivered to your Inbox.

Join 377 other followers