1945: Scientific Paper Graph Quality

Explain xkcd: It's 'cause you're dumb.
Jump to: navigation, search
Scientific Paper Graph Quality
The worst are graphs with qualitative, vaguely-labeled axes and very little actual data.
Title text: The worst are graphs with qualitative, vaguely-labeled axes and very little actual data.


Microsoft Paint was first introduced in 1985 as a component of Windows 1.0, and Microsoft PowerPoint debuted in 1990. As easy-to-use tools, these allowed for the easy creation of graphs by computer users. The comic implies that these are responsible for decreasing the overall quality of graphs in scientific papers, presumably by enabling a large number of inexperienced designers, and encouraging certain kinds of designs that are ineffective for communicating scientific results.

Critics of PowerPoint, such as Edward Tufte, have argued that the software is ill-suited for reporting scientific analyses. Many scientific journals nowadays explicitly forbid the use of PowerPoint in their instructions for authors. It can be argued that other software specifically built for this task - and techniques to do so - have been refined over time, leading to a rise in graph quality outside the PowerPoint/MSPaint era (though see discussion).

The title text states that among the bad quality graphs, the ones “with qualitative, vaguely-labeled axes and very little actual data” are the worst. While this may indicate that the problem with PowerPoint era graphs is that they seem to focus on getting the point across (qualitative as in “you get the idea”) over accuracy (little actual data), this is more hypocritical humor on Randall's part, as the comic itself features exactly that sort of lambasted graph. The vertical axis labeled “good” and “bad” is entirely qualitative, the horizontal axis manages to use numbers and still be vague by labeling the area between the ticks as decades instead of labeling the ticks, the definition of what constitutes the ‘PowerPoint / MSPaint era’ is entirely unclear, and it is doubtful that any actual data was used to make the graph – certainly there are no actual data points indicated. Its quality is doubtful, and it might represent more of an impression, or opinion, than an actual fact.


[Heading on top of the graph:]
General quality of charts and graphs in scientific papers
[A graph is shown with the y-axis on the origin labeled "bad", on the arrowhead labeled "good", and the x-axis being a timeline labeled with decades from 1950s to 2010s.]
[The pre-1993 and post-2015 parts are white, with increasing quality before 1990 and after 2015. The 1993-2015 part indicates bad quality and is highlighted in grey, labeled "PowerPoint/MSPaint era".]

comment.png add a comment! ⋅ comment.png add a topic (use sparingly)! ⋅ Icons-mini-action refresh blue.gif refresh comments!


What happened circa 2015 that marks the *end* of the PowerPoint/MSPaint era? 16:22, 22 January 2018 (UTC)

More and more journals explicitly forbade the use of powerpoint. Also, more scientists are familiar with software better suited for creating scientific graphs. Thawn (talk) 16:34, 22 January 2018 (UTC)
The problem was never that it was impossible to good quality graphs with those tools. The problem was that people didn't actually do so, in part because the tools made it really easy to produce something superficially good but actually so information-free as to be utterly bad, as well as making it rather more difficult than one would hope for to make camera-ready graphs (journals having higher-resolution print reproduction than most computer screens of the time). But before anyone gets fancy about this, you could commit very similar sins with other tools; merely using a specialist plotting program doesn't automatically make the output truly comprehensible (or relevant). 22:30, 22 January 2018 (UTC)
If, however, creating graph is harder, you are likely to focus on what to put into them and make them only if it makes sense. One reason for decreased quality of graph might be that there was more of them for same amount of data. -- Hkmaly (talk) 01:29, 23 January 2018 (UTC)
With enough effort, it is possible to make a good graph with any tool. However, the point is that with Powerpoint it is much easier to make a superficial graph than a good graph. With other tools such as R, Matlab, Origin etc. it is equally easy to make a good or a bad graph. Therefore, the average quality of graphs created with Powerpoint is much lower than with other tools. Thawn (talk) 09:36, 23 January 2018 (UTC)
I believe the point is NOT that these are bad tools for creating graphs. I believe both are capable of very fine graphs. Even Paint can be used to recreate an extremely well-defined graph which was previously drawn by hand. I believe the point is that THESE tools are not only available to but useable by scientific know-nothings. People who know or remember little to nothing of how to properly structure a graph. It's less that other / previous tools were better, more that the plebians don't know about them and/or don't know how to use them, or simply don't have access to them! Even Excel, which makes it far easier to create quality graphs, simply because it requires the actual data from which to create the graphs, and is just as available as Powerpoint, doesn't get used as much as Paint or Powerpoint, simply because they allow people to be sloppier about graphs, and I suspect most don't even know Excel can do graphs, or they don't know how. NiceGuy1 (talk) 04:38, 26 January 2018 (UTC)

An interesting thing to note is that you can see from this chart that even slightly before the paint/powerpoint era the quality started going down. But it could be because this graph is meant to be just like the point it is making and therefore is not 100% accurate. 17:47, 22 January 2018 (UTC)

I came down here to make exactly this point - Randall appears to be deliberately trying to misleadingly imply a conclusion that isn't actually supported by the data. ;o) 09:34, 23 January 2018 (UTC)
actually, the peak of the graph is somewhere around 1990 which is 5 years after the release of paint and close to the release of powerpoint. Assuming that the tools gradually went into widespread use, this is perfectly consistent. Thawn (talk) 09:36, 23 January 2018 (UTC)
Well either they quickly came into common use, in which case the labelling of the 'era' is wrong, or they didn't, in which case it doesn't explain why the decline started so early. 09:43, 23 January 2018 (UTC)
That's a false dichotomy. An era is defined by prominence not existence. There are still gas-lamps, but we are not in the gas-light era
But if there's no prominence, there should be little effect. That was kind of the point. Unless the very existence of the tools was impacting even graphs that weren't created with them. 09:50, 25 January 2018 (UTC)
It could be that it was in wide enough use to bring down the average, but not wide enough to define the era. 22:15, 25 January 2018 (UTC)
It looks to me that the downturn starts about 1988 and the era starts 1994 (i.e. close to halfway through to 90s but earlier). This would suggest that after existing for 3 years, Paint was starting to be used for enough graphs to have impact on the data (graphs created). Then, PP was around for 4 years before becoming prominent to "launch" the era where these two started being used for most graphs. Fits to me. :) NiceGuy1 (talk) 05:01, 26 January 2018 (UTC)

Does anyone have good examples of papers showing this? It would really help the explanation...

You might find http://www.norvig.com/Gettysburg/ amusing. It is the Gettysburg Address done as a PowerPoint presentation. 18:55, 22 January 2018 (UTC) Gene Wirchenko [email protected]
An xkcd PowerPoint presentation by Randall Munroe would be so fantastic! Just... don't let him near Excel: "Why does this field reference a password-protected remote SQL DB entry labeled 'Midnight Protocol' or 'else' show the time of day as a sixteen-bit floating-point decimal value from zero to one?" 'Sorry, can't hear you, headphones; I'm working on the soundtrack for the new collaborative infinitely recursive xkcd AR exhibit at Meow Wolf Ollantaytambo.'
ProphetZarquon (talk) 21:56, 23 January 2018 (UTC)
You monster! I can not delete your horrible suggestion without 1) running afoul of what a wiki stands for or 2) losing the educational value of a warning of what not to do. And then, there is 3) the perverse attraction of seeing a train wreck. 05:09, 24 January 2018 (UTC) Gene Wirchenko [email protected]
I'm just disappointed that nobody noted the 0-to-1 reference. I like trains, & I like views of stuff cutaway or disassembled, therefore a really wicked train wreck is the best of both worlds! ("Some men just want to watch the world burn. I want a soundtrack to go with it." - Me)
ProphetZarquon (talk) 18:53, 24 January 2018 (UTC)

Also amusing is how low quality the image of this comic is. It is only 360*240 pixels, which is fitting for a graph describing low quality graphs. 02:21, 23 January 2018 (UTC)

Something of which I find at least somewhat noteworthy: early xkcd was notorius for these vague, informationless graphs. 09:36, 23 January 2018 (UTC)

It resembles the uncanny valley region of the CGI humans resembling real humans. -- Comment Police (talk) (please sign your comments with ~~~~)