The authors' conclusions are overly broad given the data presented. The study subjects were tasked with copying the content and formatting of a published article. This task is not representative of actual research writing.

First, content generation for a real article is much, much slower than the time needed to copy an existing work. The 30 minutes per page that it took for study subjects to reproduce the article text is essentially irrelevant once all the time is accounted for.

Second, study subjects were tasked without reproducing formatting such as font, spacing, lines, columns, etc. (Table 1). None of these are a concern when submitting a real manuscript. In fact, publishing guidelines recommend removing all but the most simple formatting (bold, italics) from a manuscript, as the publisher will change all the other formatting regardless. It is difficult to tweak formatting in LaTeX, so the requirement to fiddle with formatting would have had a disproportionate impact on the LaTeX-using subjects. The recommended LaTeX workflow is to do semantic markup only and let the class file (either the default or the publisher-provided class file) handle the formatting.

An additional concern is the authors' invocation of a drive to reduce cognitive dissonance in order to explain LaTeX's positive usability results. In the authors' words, "This bias is usually unconscious and becomes stronger as the effort to reject the chosen alternative increases". As the effort to switch to Word is trivial in comparison to the effort to switch to LaTeX, this would suggest a bias in favor of Word rather than a bias in favor of LaTeX. Perhaps the authors meant to invoke a kind of sunk cost fallacy: users who have invested a lot of time into learning LaTeX will persist in using it so as to not render that time investment useless. In any case, considering that the study subjects are highly motivated experts in their fields and almost certainly used some kind of word processor before using LaTeX, it is more sensible to conclude that users adopted LaTeX in spite of its learning curve to solve specific problems in their writing workflow. Those who prefer Word likely never encountered those problems, making the easier-to-learn Word a more rational choice. It is unneccessary to assume delusion when rational self-interest suffices to explain the results.

Still, the article does present some interesting information:

1. LaTeX users unexpectedly (to me anyway) did better at copying equation text. Word's equation editor after the 2007 release is actually quite good and accepts LaTeX-style input, so I would expect good results in combination with the immediate feedback of the WYSIWYG interface. Which version of Word was used?

2. Copying tables in LaTeX is slow and error prone. This is probably not a surprise to anyone who has tried to edit a sizable table in LaTeX, and I suspect few people do it that way. Creating tables in in spreadsheet or statistics software, then exporting to LaTeX is easier and more accurate. Exporting tables directly from one's data analysis software has the additional benefit of making transcription or revision errors impossible (when the data changes, the table changes).

3. The high level of user satisfaction with LaTeX is somewhat surprising, given its difficulty. I suspect that this is due to self-selection: Word is the default choice, so anyone unsatisfied with LaTeX reverts to Word. Anyone unsatisfied with both Word and LaTeX (or other alternatives) simply suffers. It is plausible that in a real writing task, the lesser tiredness and frustration and increased enjoyment among LaTeX users could translate to more sustained productivity and creativity. This would be interesting material for a future study.