Thursday, May 28, 2015

Are scientists boring writers?


I was talking with an undergrad who is doing her honours project with me about the papers she’s reading, and she mentioned how difficult (or at least slow going) she’s found some of them. The papers are mostly reviews or straightforward experimental studies, but I remember feeling the same way as an undergraduate. Academic science writing uses its own language, and until you are familiar with the terms and phrases and article structure, it can be hard going. Some areas, for example theoretical papers, even have their own particular dialects (you don’t see the phrase “mean-field approximation” in widespread usage, for example). Grad school has the advantage of providing total immersion into the language, but for many students, lots of time/guidance and patience is necessary to understand the primary literature. But is science necessarily a boring language?

A recent blog piece argues that academic science writing needs to fundamentally change because it is boring, repetitive, and uninspired. And as a result, the scientific paper needs to evolve. The post quotes a biologist at University of Amsterdam, Filipe Branco dos Santos: he feels that the problem is rooted in the conservative nature of scientists, leading them to replicate the same article structure over and over again. Journals act as the gatekeeper for article style too – submission requirements enforce the inclusion of particular sections (Introduction, Methods, Results, Discussion, etc), and determine every thing from word counts, figure number, text size, and even title structure and length. Reviewers and editors are within their rights to require stylistic changes. The piece includes a few tips for better article writing: choose interesting titles, write in the active tense, use short sentences, avoid jargon, include a lay summary. It’s difficult to disagree with those points, but unfortunately the article makes no attempt to suggest what, precisely, we should be doing differently. Still, it suggests that consideration of the past, present and future of scientific writing is necessary.

One glaring issue with the post is that the argument that scientists are stuck in a pattern established hundreds of years ago ignores just how much science papers have changed, stylistically. Scientific papers are a very old phenomenon – the oldest, Philosophical Transactions of the Royal Society, was first published in 1665. The early papers were not formatted in the introduction / methods / results / discussion style of today, and were often excerpts from letters or reports.

From the first issue, “Of the New American Whale-fishing about the Bermudas” begins:

“Here follows a relation, somewhat more divertising, than the precedent Accounts; which is about the new Whale-fishing in the West Indies about the Bermudas, as it was delivered by an understanding and hardy Sea-man, who affirmed he had been at the killing work himself.”

Ecological papers written in the early 1900s are also strikingly different in style than those today. Sentences are long and complex, words like “heretofore”, “therefore”, and “thus” find frequent usage, and the language is rather flowery and descriptive.

From a paper in the Botanical Gazette in 1913, the first sentence:

“Plant geographers and climatologists have long been convinced that temperature is one of the most important conditions governing the distribution of plants and animals, but very little has as yet been accomplished toward finding out what sort of quantitative relationships may exist between the nature of floral and faunal associations and the temperature conditions that are geographically concomitant therewith.”

While this opening makes perfect sense and establishes the question to be dealt with in the paper, it probably wouldn’t make it past review without comment.

Some of my favourite examples that highlight how much ecological papers have changed come from R.H. Whittaker’s papers. He is clearly an avid (and verbose) naturalist and his papers are peppered with evocative phrases. For example, “If, for example, one stands on a viewpoint in the Southern Appalachian Mountains in the autumn, one sees a complex varicoloured mantle of vegetation covering the mountain topography” and “The student of vegetation seeks to construct systems of abstraction by which relationships in this mantle of vegetation may be comprehended.” Indeed!

Today, in contrast, academic science writing is minimalist – it is direct, focused, and clarity is prized. Sentences are typically shorter, with a single focal thought, and the aim is for a clear narrative without the peripatetic asides common in older work. These shifts in style reflect the prevailing thoughts about how to balance the role of scientific papers as a communication device versus as a contribution to the scientific record. It seems that science papers may be boring now because authors and editors would rather a paper be a little dry rather than be unclear or difficult to replicate. (Of course, some papers manage to be both boring and confusing, so this is not always successful….) Modern papers have a lot of modern bells and whistles too. The move away from physical copies of papers to pdfs and online only colour versions and supplementary information has made sharing results easier and more comprehensive than ever.

If there is going to be a revolution in academic science writing, it will probably be tied to the ongoing technological changes in science and publishing. The technology is certainly already present to make science more interactive to the reader, which might make it less boring? It is already possible to include videos or gifs in online supplements (a great example being this puppet show explaining Diversitree). More seriously, supplements can include data, computer code used for analyses or simulations, additional results. It’s possible to integrate GitHub repositories with articles tied to a paper’s analyses, or link markdown scripts for producing manuscripts. The one limitation is that these approaches is that they aren’t included in the main text and so most people never see them. It’s only a matter of time before we move towards a paper format that includes embedded elements (extending on current online versions that include links to reference papers). One could imagine plots that could be manipulated, or interactive maps, allowing you to explore the study site through satellite images of the vegetation and terrain.

Increasingly interactive papers might make it more fun to work through a paper, but a paper must stand alone without them. For me, the key to a well-written paper is that there is always a narrative or purpose to the writing. Papers should establish a focus and ensure connections between thoughts and paragraphs are always obvious to the reader. The goal is to never lose the reader in the details, because the bigger picture narrative can be read between the lines. That said, I rarely remember if a paper is boringly written: I remember the quality of the ideas and the science. I would always take a paper with interesting ideas and average writing over a stylish paper with no substance. So perhaps academic science writing is an acquired (or learned) taste, and certainly that taste could be improved, but it's clear that science writing is constantly evolving and will continue to do so.

Wednesday, May 20, 2015

I'll take 'things that have nothing to do with my research' for $400


I guess I do have a couple papers with the word fire in their titles?
And to Burns and Trauma's credit, this is a nicely formatted email and the reasons to publish with them are pretty convincing :-)

Monday, May 11, 2015

Is there a limit to how many species can the earth hold?

Counting species (bird lists, plant guides) is as old as ecology itself. And yet surprisingly, there are still different opinions on how many species the planet holds, and even, whether there are limits on how many species it can have. If the number of species has ecological limits, the assumptions ecologists often make – that species pools are limited and knowable, dynamics can reach equilibrium, competition should usually be important – would be stronger. Things would be more predictable. 

But is the production of diversity self-limited? There isn’t consensus but two recent articles in the American Naturalist (continuing a debate at the American Society of Naturalists meeting) provide some excellent debate of this question.

The debate is whether the majority of variation in continental-scale species richness is regulated by diversity-dependent feedbacks. In these papers, Dan Rabosky and Allen Hurlburt argue that species richness has ecological limits, while Luke Harmon and Susan Harrison take the contrary position, that species richness is dynamic. First, to define some terms: here, species richness is being considered at the largest spatial scale (e.g. terrestrial plants at the continental scale) so that dispersal limitation should be comparatively unimportant (because diversity changes are mostly driven by in situ speciation).

The crux of the Rabosky & Hurlburt argument is established in the Ecological Limits Hypothesis (ELH), which states that species richness will reach a dynamic (i.e. stochastic) equilibrium, where equilibrium richness reflects density dependence in speciation and/or extinction rates. Speciation and extinction rates are ultimately limited by total resource availability for the continent. Therefore variance in richness through time and between places should be driven these ecological limits, and richness should be predictable.
From Rabosky & Hurlburt 2015 - the Ecological Limit Hypothesis.
The evidence presented for the ELH comes from phylogenies and macroevolutionary models, the fossil record, and macroecological observations. First, there are well known patterns between species richness and energy, productivity, or habitat area, and these span multiple regions and groups of species (e.g. Jetz and Fine 2012). Further, Rabosky & Hurlburt argue that geological records suggest that changes in diversity are not unbounded or exponential, but instead rise and fall, correcting toward some equilibrium. Molecular phylogenies are often evaluated by looking at speciation rates over time, and the authors suggest that these frequently show declines, where speciation declines during adaptive radiations. One prediction that arises from the ELH is that perturbations will be followed by particular responses: “negative perturbations—mass extinctions, in particular—should lead to diversity recoveries. Second, positive perturbations—increases in the resource base available to a biota—predict increases in species richness to stable but greater equilibrial levels”.

The rebuttal article from Harmon & Harrison takes a strong and contrasting view, although it focuses mostly on poking holes into Rabosky & Hurlburt’s arguments, rather than laying out a competing hypothesis. If Rabosky & Hurlburt focused on evidence over huge evolutionary scales and spatial expanses, the Harmon & Harrison response has a particular interest in the temporal and spatial scales of interest to community ecologists (local, present day) and how these seem to disagree with Rabosky & Hurlburt's hypothesis.

First, Harmon & Harrison argue that that the macroevolutionary evidence (molecular phylogenies, fossil data) is not nearly so convincing as Rabosky & Hurlburt suggest. There are important limits to its utility resulting from issues of ambiguity in interpretation and methodological limitations. In addition, for most of the patterns Rabosky & Hurlburt highlight, there are other papers concluding that the pattern was not present in their data. With reference to the lack of relationship between clade age and diversity: “A common interpretation of these results is that a lack of a relationship between age and diversity is evidence for ecological limits.... However… this pattern is far from ubiquitous in real data and is compatible with other explanations”. They also take issue with the tendency for hand-wave-y interpretations of patterns in such data, and emphasize the need for better statistical analyses and consideration of alternate models. Fossil data has obvious limitations as well (hence the field of taphonomy), including the fact that fossils are rarely classified to the species-level, which means they do not represent species richness, but rather lineage richness.

But Harmon & Harrison's real disagreement is based on their view that ecological evidence from local communities does not at all suggest ecological limits. Energy-richness correlations, although common, may have alternative explanations: the tropics may have higher diversification rates for other reasons, or niche conservation means that more species niches suited to the tropics, confounding energy-diversity relationships. Further, local communities do not regularly show a positive energy-diversity relationship. In particular, Harmon & Harrison suggest that the logic from the ELH, if followed, predicts that if species richness is ultimately tied to the availability of energy, then competition should necessarily be very important in most ecological communities. They cite a stat from the invasion of California flora in which alpha diversity has risen by more than 1000 invasive species, with only 28 native extinctions (as of 2002), suggesting that local (or even regional) communities are not full. 

To this, Rabosky & Hurlburt rejoins that invasion is about dispersal changes, and not resources. Further, they believe that large evolutionary scales are most useful as evidence for the ELH, since they are most likely to show zero sum game, rather than temporary dynamics, and since confounding factors should become minimized.

The debate left me feeling a little unsatisfied (since expecting the authors solve the problem is a bit unreasonable), in part because the authors are really arguing from different scales and approaches. And both sides are clearly right in some cases (and in others, perhaps, clearly overreaching). And of course, proving whether or not there is an ecological limit on diversity is a rather difficult thing. When Harmon and Harrison argue that the ELH, which assumes that richness approaches some equilibrium value but varies about it in a stochastic fashion, isn’t parsimonious, they’re wrong – ecological processes are innately stochastic and it hardly seem un-parsimonious to assume as much. But they’re right that this view makes testing and model fitting very difficult since having high replication and good quality data is necessary (to capture accurately a distribution, rather than single value). Given the variety of issues with data representing diversification over evolutionary time, and frequently an inability to capture extinction rates with evolutionary data, having quality, replicated tests of the ELH is difficult.

On the other hand, at local scales over ecological time, observations may be less relevant. It’s not clear how to reconcile statements about saturation (or lack thereof) of local communities with richness at continental scales. Rabosky & Hurlburt suggest that local assemblages can be dynamic in diversity as long as there is a zero sum across all communities and through time. But a connection between continental scales and local scales is innate, and understanding how diversity relates over multiple spatial scales is an area of ecological research we need to continue to develop.

Given there are no easy tests of this sort of question (though bacterial microcosm provide some interesting results), we have been forced to draw conclusions based on weak tests and weak evidence. But ecologists do this because this is a truly important question, with huge implications across ecology and evolution. Ecological and evolutionary models make assumptions that implicitly or explicitly about carrying capacity, about determinants of rates of speciation and extinction, about invasion, about why global diversity changes, and these need to be confirmed. Further, if there is a strong ecological feedback of diversity, one of the most important implications is that major perturbations such as extinctions should be followed by major recoveries. In the Anthropocene, that’s an important implication.