Before asking for a text review

November 08, 2024

It is sometimes convenient, in the early stages of a project, to share a “working document”. Such a document can be “rough” and not follow the rules I listed below. However, if you are asking for an official review from a co-author, a mentor, or anyone else, you should first ensure that you can check the boxes below (note: I am sure this list will expand as I realize I forget some things). Other’s time is a valuable resource; making sure your text is as perfect as possible before asking someone else to take their time to review it shows that you understand this and respect this person’s time. The following list applies to drafts for papers, proposals, theses, dissertations, grant proposals, etc.

  • The final version of your text should have been run through a grammar and spelling corrector. I recommend Grammarly.  This website provides a score. Most of what I submit for publication is between 98-100. Some more technical writing may be hard to bring above 95, but you should strive to do so. If your text is below 90, you need to change something (eventually, your writing style!). Personally, I am fine working with Grammarly’s free version. It does not give explanations or suggestions for everything, but it underlines where there are some issues. I work on these segments until they stop being underlined. It is often for the same reasons. For example, we tend to use a lot of passive sentences (“EEG was recorded.”) instead of active sentences (“We recorded EEG.”) which is generally discouraged. Please use Grammarly systematically or an alternative product if you have an alternative that works better for you. If your text scores below 95, I will likely return it to you without review. For reference, the text of this post is scored 95.
  • Make sure all your figures are high-quality. For that, please follow the instructions in this post.
  • Make sure all your claims are supported by 1) proper citations or 2) statistical testing. Note that a classification system with a 95% accuracy does not “outperform” a system with 94% accuracy unless you can demonstrate that this difference in accuracy is statistically significant. Examples of approaches to do that include bootstrapping your accuracy assessment. This comment is true for ALL such results. Point estimates do not allow for statistically valid conclusions; only distribution of results can allow for such a conclusion because they allow assessing the reliability of an estimate. 
  • If you worked with Latex, you must proofread the PDF. The PDF rendering, not the Latex code, needs to be flawless.
  • Did you re-read your text at least (again, at least) three times? I re-read mine many more times than that. Every time I do, I change a lot of stuff. If you did not re-read your text at least three times, the quality of the write-up is likely poor. One re-read is fine for emails and unofficial communications. You should proofread anything else with much more care.

Do you need more help to learn how to write excellent texts? USC has a writing center to help you with that.