A nice article on software patents by Joel Spolsky has an intriguing paragraph that may shed light on a little-noticed problem with the peer review system:
The first technique is to try to make the language of the patent as confusing and obfuscated as possible. That actually makes it harder for a patent examiner to identify prior art or evaluate if the invention is obvious.
For similar reasons, it is in the best interests of the researcher to make their peer reviewed article as incomprehensible as possible.
When confronted with closely reasoned and heavily mathematical text, the reviewer is presumed to have been chosen to be sufficiently familiar with the field as to be able to follow the reasoning without difficulty.
But consider the position of the reviewer who does not understand the argument in the paper under review. While there are degrees and shades here, fundamentally the reviewer is choosing between two poles: 1) I am a fraud, as this material is clearly on my territory and I don’t understand it 2) I can’t understand this because it is completely incoherent gobbledy-gook done so effectively that I can’t immediately see what is happening here. The reviewer could simply refuse to review the paper “on reviewing the paper I find I am not qualified to give an opinion” which solves the reviewer’s problem but not the editor’s. The editor has to find three reviewers to complete the process. In any case this paper is likely to sit on the reviewer’s desk for a considerable time; and then the reviewer will feel pressure to complete the forms and get the task dispatched with.
And as with the less qualified person on your PhD committee, what they will go with is likely to be “I don’t see anything wrong with the parts I am in a position to judge” and give a pass. On the other hand, if a paper has plain, simple unobfuscated language and a frank assessment of its own weaknesses, then the overwhelmed reviewer is in a position to give it a good, solid bikeshedding without much effort. Since career output is measured by publication quantity first and quality only second (indirectly via citation count), clarity and completeness are not encouraged.
I lack for a proper adjective to characterize the resulting trends in the literature, but “obfuscated” will do for an understatement.
I’m not trying to say that there aren’t fundamental reasons that many papers are only accessible to peers; good science really is deep, and it’s ludicrous to propose that everything be accessible to a lay audience – that would make most genuinely important papers thousands of pages long. But that just provides cover for the more pernicious phenomenon that I am suggesting is also at work. Papers should be as readable as possible to their intended audience on principle; in practice they aren’t.