Standard Error or Standard Deviation?

I was surprised when a colleague asked me for a opinion on a revision of a paper mentioning they should present Standard Deviation instead of Standard Error. I went to the website of the Psychiatry Research, Guide to Authors, and there it was the recommendation to use SD instead of SE. It is as dry and enigmatic as this "Study group variability should be reported as the standard deviation, not the standard error". I am not sure what they mean.

I am not against using SD when it is appropriate and would not discuss the fact the many folks may not understand the difference or even use them incorrectly. But I think a general request to use SD is not reasonable, a more elaborated instruction should be given, perhaps referencing some paper since there are several talking about the difference between SD and SE. It turns out that in that specific case I totally disagreed with presenting the SD because they where presenting tables mostly with %, comparing these % across three groups. Even though the tables were preliminary descriptive analysis preparing the reader with some background knowledge about the variables that  were later included in the model. SD is next to useless for %, really, and the SE would provide the reader with a quick idea of whether the group means differed and how precise they were.

And this incident followed a rejection by another Journal, a week earlier, of a paper I thought was very good. There were 4 justifications for the rejections and all they did was to show that whoever wrote that did not understand anything about our model. The first thing I commented with my collaborator was "How can a person review something they don't know anything about?" It seemed blatantly unethical to me, it seemed so not science.

Then you cannot say much. It is like they are Godlike beings, it is like they know more than you, period. I think the system for publication and peer reviewing needs to be different. I would not mind talking to them about our paper, maybe a phone call, maybe talk to their statistician if they have one.

I decided to take e a quick look at the Psychiatric Research papers, I wanted to see if they only publish standard deviation. I went to the first paper of the last issue (well, this was the last available to me). In session 2.2 Subjects they do present some SDs and I think that is appropriated because they are describing the sample, not making inference. However, the way they present, with that plus/minus sign, makes things confusing. It looks like a confidence interval, which is inferential procedure and therefore makes us think they are presenting SEs (or 1.96*SE). I always found this plus/minus notation very confusing not to say misleading or plain wrong.

Then we have Table 3, where 4 groups are being compared in a inferential way, there is a ANOVA there. Then they show mean and SD? This is where I disagree. It would be important to have SE here, maybe confidence intervals. I think here we are talking about means and we want to know how precise they are, that is, the SEs.

So, I don't know, I think this instruction to use SD is already causing more harm then good because there is nothing like use this or that, different things are appropriated for different occasions.