Figures...
Sunday, December 11, 2011 at 11:46AM I think that I've begun classifying different sub-disciplines within the biological sciences based on how obsessed they are with manuscript figures. An oversimplified continuum may place 'figures are the only thing that matters' on one end, versus 'actually some things are better explained by tables and text' on the other1. Geneticists? Very far on the figures side. Evolutionary biologists? Text-of-center. Population geneticists? I suppose that a plot will illustrate the power of my math to the savages. And so on...
In all seriousness; coming from a population genetics/evolutionary biology background, I certainly recognize the value of illustrative figures in conveying a point. However, I have noticed that there is a massive difference between my field and the more figure-obsessed disciplines: In genetics (for example), the figures are the paper's story, and the text is written in order to explain them. In evolutionary biology/population genetics, the figures are produced in order to illustrate the results of an often computationally complex analysis of multi-dimensional data.
While there are a few tweaks that you can pull in terms of how to best present a western blot, there are often many, many different ways to present the results of a clustering algorithm, or to represent a gene expression profile among other examples. Perhaps I'm just not very good at this stuff, but I've found that it's incredibly difficult to choose the most appropriate representation of the data a priori. Rather, I find it necessary to flesh out the overall thrust of a manuscript using place-holder representations, and then choose the best way to represent the data once it has all come together.
This is why it was a major sticking-point for me when a former boss used to say that he wasn't interested in seeing any manuscript text until he'd seen the finished, publication-ready figures. I wouldn't subscribe to this approach in general, but I can see how this would be much more 'do-able' in certain fields as compared to others. In my case, it was rather paralyzing.
Here's another thinking point: I've also learned that some figure-focused people only look at the pictures when they read manuscripts. I'm not entirely sure that this is a good idea in any field, but it seems especially potentially misleading in large-scale datasets: Before I accept the validity of an analysis, I want to know that the sample sizes were appropriate, the null hypothesis was warranted, the statistics employed fit the data, etc. In evolutionary biology, these factors can change a solidly supported model to pure hand waiving speculation. If you only look at the figures, you'll rarely be able to distinguish one from the other.
1Alright, a more 'extreme' continuum may end with, 'I'll just crap the figures out in MS Paint the morning that the paper is submitted'. However, while I have actually witnessed the 'FIGURES ARE EVERYTHING' school, I haven't seen anyone who didn't care about them.
Reader Comments (5)
Oh, it is so easy to make a figure lie! You can play with the scale, with the axes, with the colors. And a nice regression line will make any cloud of data points look interestingly correlated. A good figure can conceal more than a thousand words.
Heh good one. I have actually read papers where I thought the figure might actually be the bane of the paper itself. When you look at it and you're thinking in your head - I'm interpreting that entirely differently. I also agree with what the other person here just said about scale. I have read papers where a small decline was made to look massive by simply changing the scale.
A month ago or so I testified before parliament. One of the major difficulties I had was presenting the data in such as way as a layperson could understand multivariate regressions. It felt odd presenting something in a manner that one would never present it to other scientists, but the point was a lot easier to make when the MP's saw the graphs.
There's no doubt that a good figure can be use to 'illustrate' a point more forcefully. It's also quite clear that figures can illustrate mistakes, as is the case where people draw regression lines through data that have obvious oddities in structure (reduced variance at one one end of the axis, for example).
I've just been a bit stunned when I speak to people who say things like: You should be able to interpret the entire paper via the figures, which just doesn't hold water in my field. We did a journal club paper a few weeks ago where the figures drew an elaborate model that the quality and quantity of data didn't support at all. Without reading the text, I suppose you have to assume how things were done and that they were done correctly (or at least ignore major potential caveats).
I decided to procrastinate a bit and read the most recent update here.
I'm procrastinating working with the figures for a manuscript that is my number one priority (high level procrastination, in other words).
I'm not sure if this is somehow contributing to anything right now. Sorry, brain full of scatterplots.
Ah, a very good observation! I come from the opposite direction (cell biology to evol/popgen), so I'm one of those people who looks at figures first, and usually only read parts of text to explain them. In cell biology, our data ARE figures, so I guess it makes sense, as it's a freaking nightmare to try to convert micrographs into text and vice versa when reading that. My old boss has been pushing me to work on our manuscript... by asking for figures. Which of my micrographs are suitable for publication? How do they fit in our story? Basically, the figures write the paper, and the text just holds them together and provides background info.
As you point out, life is very different in evolutionary biology. Suddenly, people give me papers with NO figures to read, and as a slow and crappy reader... it hurts a little ;-)
I also blog around figures; the story is written to fit the images I have, preferably nice ones. Normal people probably use figures to support their text instead...
But meh, this is why I'm obsessed with microscopy, and only do quantitative and verbal work when beat with a stick. I like pictures! Me abstract logic iz no good =/