Friday, October 3, 2014
An Ode to the Review Paper
Each paper yields just one or two scraps of relevant information; information that is often incomplete or contradicts that of other papers. Half the time you can't even judge if the data will be relevant to your work. There's nothing surprising about this, of course. The low-hanging fruit ("neurons use electro-chemical signals!", "the eye is a lens!") have long since been picked, and it takes a lot of time, ingenuity and effort to tease out even a few new scraps of knowledge in detail. Most papers present only a few small new findings because that's all a year or more of work will yield.
But it means any search for background information is a long, slow struggle. And it never ends; there's always more to find if you look. More than a few graduate students get caught in this trap: they think they need all the background before they do their own work, then get stuck in an endless loop of literature searches with no results of their own to show as their grad school clock runs down.
Enter the review paper. That's a paper where a researcher collates, summarizes and interprets all the recent literature on a topic. It can run to dozens of pages and contain hundreds of references. It summarizes the current state of knowledge, and is a great one-stop source for all the major papers. A good review paper also highlights areas of disagreement and areas we don't know enough about.
Some people criticise review papers. They tend to get a lot of citations, often more than the original papers, even though the author often hasn't done much or any of the work it presents. But I say review paper authors are if anything receiving too little credit, not too much.
It truly is a monumental struggle to go through and synthesize all that material. A well-written review paper is a tremendous service to the field, a time saver for hundreds of other people, and a great way to make clear what we do and do not actually know. The more we have of them the better.