When not to read the literature

That’s the shelf in my office where I pile new journals still in their mailers. It’s embarrassing for two reasons.  First, it betrays my age that I still subscribe to print journals.  Second, there must be 2 years worth of unread issues in that pile!

Looking at the pile every day led me to think about my literature-reading habits, and then to ask myself a question: when should one read the literature, and when should one not? My answer: one should generally not read the relevant literature before designing and executing a piece of research.

That advice probably seems surprising. After all, if you read any paper (or proposal) I’ve ever written, you’ll find an Introduction that reviews the literature background, points out a knowledge gap or a point of contention based on that literature, and then explains how my research was carefully designed to fill that gap or address that controversy. This is such a familiar and useful convention that it can be hard to imagine doing science any other way. And yet in many cases, it’s not actually what I do. I think that often, too much familiarity with the literature makes our science worse rather than better. Overfamiliarity can lead to two somewhat different kinds of problem: first, it can seduce you into accepting a field’s dominant paradigm rather than questioning it; and second, it can fool you into thinking a question is more settled than it actually is. I’ll illustrate each with an example from my own work.

In a previous incarnation as a stream ecologist, I needed to know how quickly material from a near-point source gets mixed across the width of a stream channel (never mind why). We did some tracer experiments to find out, and found that mixing takes a long time: tens to hundreds of metres, even in small streams only a metre or so across. If I had read the stream literature before heading to the field, I wouldn’t have done this work, because I would have learned that small, shallow streams are turbulent and therefore well mixed. Or at least, I would have “learned” that, even though it isn’t really true – it’s just part of the dominant paradigm in stream ecology. (By the way, slow mixing is well known, and spectacular, for confluences of large rivers.)

A more important example came later, when I was first dipping my toe into studies of host-race formation in phytophagous insects. I knew that the goldenrod ball-gall fly (Eurosta solidaginis) had host races on a pair of goldenrod species, thanks to years of great work by Warren Abrahamson and his colleagues. I wondered if the same might be true of the goldenrod spindle-gall moth (Gnorimoschema gallaesolidaginis), another gallmaker attacking the same pair of goldenrod hosts. So I talked my friend and colleague John Nason into collaborating on some population genetic work, and we found that the spindle-gall moth is a pair of cryptic species. Only later did I discover a monograph in which Bill Miller had already described moths from the two goldenrods as distinct species (on somewhat arguable morphological grounds). Had I read that first, I would have ticked the question off as answered, and moved on with my curiosity satisfied. Instead, we built our spindle-gall moth result into a productive run of papers. For example, we’ve shown that at least three gallmakers have evolved races or cryptic species on the same pair of goldenrods, via host shifts that occurred in the same direction but not at the same time. More recently, we’ve found that the spindle-gall moth has a greater impact on its evolutionarily novel host, as a parasite might. These things are interesting and important, and reading the literature could well have derailed my path to learning them.

To be clear, I’m not suggesting you never read the literature. If nothing else, after you’ve done some science, you need to discover related literature so your papers can explain how your results confirm, extend, or challenge current knowledge. More broadly, of course you need to know enough about a field to ask sensible questions and be aware of feasible ways to answer them. This is especially true when you shift into a new research area.  When a question really has been thoroughly thrashed out, it helps to recognize that so you can find a loose end to tug on, and avoid repeating past mistakes while doing so. Finally, the literature can be a great source of ideas for future work – especially (maybe only?) when you’re reading papers about systems or questions different from those you’re working on. So don’t ignore the literature completely – just keep in mind that reading too much of it too early, and reading things too close to your own work, can sometimes be a straitjacket.

But my pile of journals is still embarrassing.

© Stephen Heard (sheard@unb.ca) Jan 7 2015

---

UPDATE: Since writing this, I re-read Peter Medawar’s (1981) “Advice to a Young Scientist”.  Among his sage proclamations: “Too much book learning may crab and confine the imagination”, and “The beginner must read, but intently and choosily and not too much”.  Good to know I’m echoing a Nobel laureate, but I’m not unaware of the irony of ignoring the literature on when not to read the literature…

---

---

Advertisement