[Update: If you like this, you may also be interested in the post I did on turning the same material into an academic talk.]
What follows is the process, from beginning to end, of creating a poster for an academic conference (in this case, the Genetics Society of AustralAsia conference in Melbourne which begins this weekend). It should be made clear: this is happening right now. It’s science that I’m just starting to write up for publication, at the same time that i present it as a poster at the GSA and as a talk at ISBE next month. I’m posting this for several reasons: the hope that it helps someone in need, the entertainment value from my struggles, and a look inside how this part of science gets done.
Note: what you’re getting here is a peek inside a part of my process for doing science. There’s a few things it’s not. It’s not a finished product like the paper that we’re going to write will be. It’s not without warts, or hiccups, or half-baked ideas. And this post is not about the content as much as it is the design. It’s is a series of snapshots during my design of the poster, meant to show you a bit of how I do these things. I’m continuing to think about and refine the ideas in the poster, and I’ll be taking into account feedback from the conference and from people I show this to. So, if you show up and bleat out “lol thats stupid y u do tht?”, I’ll slap you. Having said that, questions about the process or the material are more than welcome.
(Update, August 7: Thanks to a comment by Stuart below, I’ve realised that I never specified what tools I’m using here. I created this poster on a Mac using Pages to do the layout, with ColorSchemer to help with palette selection, Pixelmator to deal with bitmapped images, LaTeXiT to insert equations, and Omnigraffle Pro to produce vector graphics like the model diagram and the niche construction explanation. The graphs were created in R (using ggplot) and the model dynamics plot was done in Mathematica).
– Thursday, 5:00 p.m. I’ve just had a meeting with Mark, my boss, discussing the results for the model we’ve been working on. This is the model that will form the basis for the paper. I’ve still got work to do, but it’s done enough that I can begin laying out the poster.
– 6:30 p.m. I finish up the first version of the layout.
Click the image to link to a .PDf version.
Note that I’m following, at least vaguely, the design philosophy of separating presentation from content. It proceeds iteratively, and I’m not showing you all of the steps, but even in this early version you can see that I’ve blocked out an outline of what the content should be and I’m creating the aesthetics in a parallel process. To see what I mean, imagine replacing all of my text in this image with your own content, and you’ll see that it’s not too difficult. You might need to add or subtract a box, for instance, but the general layout is flexible enough to contain your content no matter what it happens to be. This separation of content and presentation is behind a lot of the best publishing technologies; it’s why I use LaTeX and not Word, it’s why we have HTML (and friends) and CSS , and so on.
– 6:35 p.m. I begin to play around with the color scheme a bit. I’m mucking about with color palettes from COLOURLovers and importing them to Colour Scheme Studio, then exporting the palettes to a .clr file and importing them into OS X’s Color Picker (open the Colour Palettes tab on the picker, click the gear and select Open to import). For giggles, I’m playing around with a rather pseudo-Satanic palette called ‘Red Devil’.
Click for PDF
It’s not really working for me, but that’s okay. I’m planning on keying the poster to my graphs anyways, by using ColorSchemer’s awesome PhotoSchemer feature to construct a palette from an image of my graphs. Of course, that means that I need to produce the final graphs first and I haven’t done that yet, so it’s back to the grindstone.
– 6:57 p.m. I send the black and white layout to my wife, who’s in New Zealand at the moment. Her first degree is in fine arts and visual design, and she’s taught me almost everything I know about design, so I value her opinion greatly when I’m working on a project (you should have seen the face she pulled when I showed her my first-ever poster as an undergrad). She’s okay with the layout, but doesn’t like the header font. This confirms my niggling doubt that the font isn’t a great choice, so I’ll replace it in the second layout tomorrow.
– Friday, all afternoon. I’ve been using Mathematica to analyse the model, but Mathematica’s graphics are terrible, so I’m reimplementing everything in R.
– Sautrday 5:14 p.m. I’ve arrived back in the lab after spending the day running errands. I was up until 3:30 a.m. last night working on the R code to reimplement all of my plots from Mathematica so that I can make more flexible and aesthetically pleasing graphs. It took a significant amount of trial and error, but it was worth it. Now, I can start thinking about the design again. I’m starting with colours: the palette I used while creating the R code is here, and I also like this variation. My concern is that it might not be bright enough to catch attention in the poster session, but I’ll see how it looks when I try it in the layout.
– 5:48 p.m. I’m down the rabbit hole looking at colours.
Looking at the graphs, I’m worried that the gradient fill is more visually pleasuring and would be easier to design the rest of the poster around, but in terms of reading the graphs I’m thinking that more contrast in hue among the values will aid the viewer in picking out interesting values. I ask my wife via text in New Zealand and she replies with something awesome: “I prefer the first one, but if the second one reads more clearly, go for that. Form is important, but function is non-negotiable.” This is a great reminder of an important point. While it’s true that too many scientists ignore form completely, function still comes first.
– 8:15 p.m. I’m still tweaking the first set of model plots. And yes, it’s 8:15 p.m. on a Saturday with me still in the lab. Don’t you judge me.
– 8:16 p.m. To hell with it, I could be doing this for days. It’s time to iterate the poster layout. First, I need to find a new font for the header: to the internets! I wish that I was experienced enough to have a set of fonts as go-tos for this, or to be able to think of the right font at first go. But I’m not. So instead, I leverage Google and troll for suggestions.
– 8:30 p.m. I’m trolling fontsquirrel and Google for fonts, and I’ve picked a few to try out. Based on this random post, I hit on a couple of combinations of font families that don’t look too bad.
PT Serif and Sans
– 9:33 p.m. The PT Serif and Sans have a nice clean feeling to them, and I’m leaning this way right now. I’ve emailed the samples to my wife for further advice, but she’s gone to bed so I’ll hear back in the morning. If you’re keeping track, I just spent over an hour playing with fonts. This either means that I have (a) great attention to detail, or (b) goddamn OCD. It’s getting late and I’m getting burnt out, so I’m going to work on this for about another half hour and then finally get out of here.
– 10:25 p.m. After spending a while fixing further problems with the plots (making backgrounds transparent, etc)., I’ve messed around with the design somewhat. It’s still early days, but a few of the elements are there.
Click for the .PDF
I’ve included one graph in each color scheme to see what works best, but I’ll make a final decision later. Also, one thing I notice immediately is that the author text in the top right now looks a little awkward. I’ve created a column line that the author text is crawling out of, which looks jarring, so I provisionally shift it left.
– 10:36 p.m. Am I seriously still in the lab? It’s time to go home.
– 10:44 p.m. Damn OCD. I realise that I’m still playing with the layout, seeing what gradient fills do. I’m not committing to these at this point, but restrained usage might be helpful. Also, I’ve added drop shadows to the colour boxes to separate them from the page a bit.
Click for .PDF.
Now I’m definitely going home.
– 10:53 p.m. Okay, for crying out loud, I’m walking out the door now.
– Sunday, 4:49 p.m. Back in the lab. Just shut up.
– 5:36 p.m. I can’t find a good vector example of the niche construction diagram that – to me – encapsulates the idea so well. Even the ones in the published papers are, usually, low resolution or have bizarre stippling. Thus, I’m replicating it by hand in Omnigraffle. On the upside, I can then sync the diagram’s font to the font in my poster.
– 6:26 p.m. I’ve recreated the diagram and placed it on the poster. I’m now beginning to fill in the text, and I run into an immediate question: what do I do with references? In previous posters I’ve used the standard in-text citations and reference list at the end, but for this one I have the feeling that space is going to be at a premium. I can’t afford to throw reference list in anywhere here, so what to do? I consult an oracle, the Better Posters blog, and on the advice to get away with what you can, I decide to play with QR codes and/or a separate printed reference list to be handed out.
-7:27 p.m. Unsurprisingly, I have too much text in the first text box. I’ll probably spend the next hour trimming words to economize as much as possible. Every word I can get rid of now without diluting the message makes my life easier down the road.
Click for .PDF
– 8:39 p.m. I’ve just sent an email to the authors of the paper whose work started my interest in this topic, asking for permission to use one of their images in my poster and the talk I plan on giving at ISBE. Get permission for your images!
– 9:00 p.m. I’m struggling with the text. I’m beginning to think that I’m going to have to omit the phylogenetic work and leave that for the talk and paper.
– 10:48 p.m. I’m still struggling. I’ve put in the model details, but they simply take up too much space.
Click for .PDF
I ended up dumping one of the gradient boxes to try and fit more of it in (and because it was difficult to get the equations into white to show up on the box’s fill; not impossible, but annoying). Still, I’ve got too much. I think that I’ll have to cut the equations; they’re effectively redundant anyways, since the model diagram actually contains all of the same information. I wish that I could cut the parameter explanations, but I don’t think that I can afford to do that. They’re just incredibly awkward and space-consuming.
– 11:11 p.m. Okay, if I don’t leave the buses are going to stop running. But in the intervening time, I did rearrange the content somewhat. It’s still ugly as sin, but it’s starting to approach something useful.
Click for .PDF
– Monday, 12:05 p.m. Had to pick my wife up from the airport this morning, but now I’m back in the lab.
– 3:25 p.m. Just got back a half hour ago from a talk by Paul Davies, which took up a good portion of my afternoon. On the upside, I’ve finally managed to squeeze everything in the left column, even if it’s still hideously cramped. I need more negative space, Captain!
– 5:30 p.m. I had such high hopes for this afternoon, but I have to swap out one of the plots and of course Mathematica is taking forever.
– 5:49 p.m. I promised my wife I wouldn’t work late in the lab when she got back, so I’m headed home. This is as far as I’ve gotten today:
Click for .PDF
– 10:13 p.m. My wife has gone to sleep; time to spend a few hours working on this thing!
– 12:37 p.m. The first semi-complete version is done! I need to finish off the conclusion sentences, fix a plot background to be transparent, and make about a thousand other tweaks, but I’m inching closer. I’m using colour in the headline now to call out key words (viral niche construction, and accenting our last names). You’ll also note that the QR code for references makes its first appearance.
Click for .PDF
– Tuesday, 11:09 a.m. I’ve been working for about a half hour now, making small tweaks and fixing the text. (How tired was I that I equated virulence with R0)? I caught Mark – my boss – in the hall and worked out a quid pro quo; he’s agreed to look at my poster and help fix the text if I give him some new figures for his talk. So now I’m going to print it off in its current state, give it to him, and concentrate on fixing the rest of the design. We’re entering 10% territory here.
– 11:26 a.m. Of course the printer is borked. Of course.
– 12:35 p.m. Am I really still trying to get Mathematica to save this PDF properly?
– 1:21 p.m. I give up. Everything I try fails: Mathematica continues to clip my legend so that the last few characters are cut off. It looks fine on the screen, but the .PDF is clipped. I have to admit defeat and abbreviate the labels. It looks terrible, but there’s apparently nothing that I can do and I’ve already wasted an hour and a half of my life on this. As I put it on Twitter, very eloquently:
– 2:23 p.m. Mark gave me comments on the text and there’s only minor changes to make, so I can finish those tonight and actually get this thing done on time!
– 10:57 p.m. After a few final changes, a spell-check, a few mutters and curses and tiny tweaks, the final version is ready!
Click for .PDF
Conclusions and lessons learnt: I’m still somewhat disappointed with this layout, because I had wanted to do something more radically simple. But there are constraints when doing this sort of work, which I need to learn to be okay with; the fact is that some forms of content can only be simplified so far before the message begins to bleed away. That’s not to say that I think I succeeded here, but rather that I did the best I could and I’ll continue to search for ways to make it better. In the meantime, I think that the layout is functional if not fantastic, and I hope that it presents the deluge of information in as straightforward a way as possible. There’s a lot of quibbles you might make about my choices: my choice to abandon headings (‘intro’, ‘results’, ‘conclusions’) in favour of a straight read-through, colour choices and design elements, using a QR code for the reference list (I’ll have an envelope with paper copies of the same list as well, but still). If you have any questions or comments, please feel free to leave a comment or hit me up on Twitter.
Addendum: Should I be surprised that when I went to go print the poster the next day, the plot in the top left disappeared from the .PDF I submitted when viewed on a Windows machine? I had tested it on several Macs to make sure it worked properly, but we don’t have a Windows system around so I didn’t see the problem. I spent about two hours tracing it to a problem with – you guessed it – Mathematica. For some reason, the .PDF produced by Mathematica just disappeared when viewed on Windows. I couldn’t fix the problem directly, but if I exported from Mathematica to EPS and then had Pages import the EPS (forcing it to generate a .PDF of its own to place in the layout), it finally worked properly. As of now, the poster is sitting on my desk, ready to go when I get on the plane tomorrow. Wish me luck!