I cannot seem to find anywhere in the paper that describes whether the mistakes were counted during or after the participant claimed to be finished. If it's the latter, then it could easily be attributed to the lack of a live preview and spell-checking, which is a technological problem easily fixed.

Moreover, the study only uses extremely small documents (with only a few mathematical equations) in their study. The power in TeX comes from defining project-wide macros and label management, allowing for instantaneous refactoring, tidy management of multiple versions of a document using imports and versioning, and seamless integration with whatever terminal tools you already use.

In other words, this paper's implied conclusions are short-sighted because TeX does much more than print the written word (or mathematical/tabular data). I would like to see the same study extended; make the same groups refactor a twenty page document, replacing one notation with another, reorganizing the sections, adding new figures, and reformatting the bibliography according to a different standard.