Whilst browsing the writeLaTeX twitter feed the other day, I came across an interesting observation that the absence of TeX is a good indicator that a math paper has issues, highlighted by John Cook (@TeXtip) in this tweet:

Not using TeX is the first sign that a math paper might be wrong: http://t.co/Q6CO8YFID7

— TeX tips (@TeXtip) July 19, 2013

In the original blog post on this (entitled ‘Ten Signs a Claimed Mathematical Breakthrough is Wrong’), this is only one of a number of points made, although it is the first, and is accompanied by the interesting statistic that:

“This simple test (suggested by Dave Bacon) already catches at least 60% of wrong mathematical breakthroughs.”

This post started me thinking — when working in mathematical physics research, I always used the arXiv as my main source of papers (new and past), and whilst I would never mind if a paper only had a pdf version available (as opposed to the full source), I don’t recall ever coming across one that didn’t look like it had been produced by TeX. So in that sense it’s hard to judge.

However, when I moved to transport research, LaTeX is the exception rather than the rule. This naturally makes it much harder (if not impossible) to make any judgement simply based on whether an author has used TeX or not (as an aside, one of the reasons writeLaTeX came into being was to enable collaboration on transport research papers!). However, I tended to find far more errors and inconsistencies in equations in the non-TeX papers, which made them harder to follow, but by no means less important or less correct in their actual conclusions.

Looking back — I think a lot comes down to how anyone with expertise in a particular field makes judgements based on their experience.

In the specific fields of maths & mathematical physics, I would have found it surprising to see a non-TeX paper, which would probably mean it would have been pushed lower down my priority list to read….and given the volume of papers there are, it almost certainly wouldn’t have been read unless there was some other reason (e.g. a recommendation from a colleague). So if you want to be read and are publishing in the fields where LaTeX is established as the norm, I think it’s important to make sure your paper has a TeX (or equivalent) quality of typesetting.

In other fields, I think it’s a much fuzzier line, and is more an indication of the background of the author than any indication of value of a paper’s conclusions. I personally find that a paper which has – on first glance – a set of clear diagrams, a clear presentation of results, and a well structured conclusion is most likely to grab my full attention.

But then again, this is what you get with LaTeX, right? 🙂

PS: I reposted John’s original tweet to our various social media sites, and the mathematics community on G+ generated some interesting follow up comments (including my original thoughts which formed the basis for this post).

Harvey Friedman often writes his papers not in TeX-based format, at least for a first go: http://www.math.osu.edu/~friedman.8/manuscripts.html

Hi David — that is a good point, and there are definitely some good counter-examples (such as the two mentioned in the original blog). The article just made me think of my own experiences, and how they were very different in maths vs transport research.

This is also a good opportunity to point out that for data visualization in articles, Edward Tufte has produced some great examples and guides (and inspired two LaTeX document classes — http://www.ctan.org/pkg/tufte-latex)

The way I’d phrase it is that it’s not logical implication but conditional probability:

Pr( correct | not written in TeX ) < Pr( correct| written in TeX )

If someone writes a math paper without using TeX, that's evidence — not proof, but evidence — that they're an outsider. And it is unlikely — not impossible, but unlikely — that a paper on a long-standing math conjecture written by an outsider will be correct. These are not logical necessities but statements of conditional probability.

When the post said that the TeX test "catches at least 60% of wrong mathematical breakthroughs" I imagine they're saying that Pr( not TeX | wrong ) = 0.60. I would expect the reversed probability, Pr( wrong | not TeX ), to be near 1. Pr( wrong | TeX ) is fairly high for major breakthroughs, and it would only be higher given something is not written in TeX.