Vivek Haldar

Team Science

The era of the single-author paper is over, and has been for a while1. Especially in the sciences, where the trend towards longer author lists is strongest, and accelerating the fastest. Physics is at the forefront, with the first thousand-author paper recently published.

This is happening without any loss of impact, i.e. team-authored papers have been getting more citations than their solo-authored counterparts, and the gap is widening.

Extrapolating current trends, there will be no new single-author papers in the sciences by approximately 2020.

The data only confirms what most people in the sciences already knew for a long time. It is now completely a team activity. Each player is incredibly specialized. So much so that to write a single paper, which itself is about something highly specialized, requires a team of authors, each of whom has specialized in one little part of that. You could be grinding away in a lab for months, doing experiments, generating data and analyzing it, and you could end up contributing only a few paragraphs.

Not even half a lifetime ago, the giants of science were comparable in stature to the size of their fields. But now, they are more like generals commanding vast armies to climb uphill. They might have the authority and the vision, but nobody, including themselves, believes even for a moment that they could have made their conquests without their armies.

Special/General

So should one specialize or generalize? It’s a question of degree, but one can still pick a direction: to know more and more about less and less, or to sacrifice some depth to gain breadth.

In at least one study2, marked gains have been found for specialization. But the study was done done in the context of pro football, and I’m not sure about its applicability to knowledge work.

However, all the macro-level forces are pushing researchers towards specialization. The complexity simply demands it.

But then who has the bigger picture? Whose hands are on the rudder of this giant ship of worker bees? More often than not, that person is a technocrat–a cross between a scientist and a bureaucrat. They are smart and technical enough to understand the work. They are good managers. They are great at selling the work to funding agencies. They are the ones keeping the lights on.

Just like in almost every other field of human endeavor, the middle is gradually being hollowed out. You start out as a lab rat. If you are really good, you will eventually become a technocrat. But if you are only mediocre, you will never be able to jump across the vast chasm between the two.

This is where deep specialization might actually end up hurting you. Because in order to be a technocrat you need to generalize. You need to know how your tiny little field fits into the bigger jigsaw puzzle. You need to convey this to the people with their hands on the purse strings.

HAL, please list the experiments to be carried out

At the end of Kevin Kelly’s talk3 about how science is carried out, someone asked an interesting question: “so we’ve already seen the first thousand-author paper. When do you think we’ll see the first zero-author paper?”

The answer to that question will probably define the nuts and bolts of how science is actually carried out in the future. If the totality of the knowledge in even a highly specialized field is beyond the comprehension of not just one but even a small team of humans, and if all they can hope to do is make a long series of small edits to the giant wiki of science4, will we eventually turn to machines to summarize that knowledge into terms that can be readily understood, or going even further, suggest fruitful opportunities for investigation?


  1. The Increasing Dominance of Teams in Production of Knowledge, Stefan Wuchty, et al. Science 316, 1036 (2007) ↩︎

  2. Does it pay to specialize? The story from the Gridiron, Rob Simmons and David Berri, Lancaster University Management School Working Paper 2007/044 ↩︎

  3. Kevin Kelly, The Next 100 Years of Science: Long-term Trends in the Scientific Method. ↩︎

  4. Repositories of science ↩︎