Are physicists hot or not?

It has become trendy to study scientists. Two weeks ago, a group of network researchers published a paper in “Scientific Reports” that aims to analyze in how far scientists pay attention to what is trendy.

Do scientists trace hot topics?
Scientific Reports 3, 2207, (2013).

The title is however misleading for several reasons.

The most obvious reason is that the analysis presented in the paper was performed exclusively on papers published in the Physical Review journals (in the years 1976-2009), meaning the word ‘scientists’ would better be replaced with ‘physicists’. Even that would be misleading though, because it’s questionable that papers published in the Physical Review are representative for the whole of physics. Physical Review is a high quality journal and it tends to be conservative. If your research is speculative or on a highly specialized topic then it might not be your journal of choice, or so a friendly editor will write before marking your manuscript as “no longer under consideration.” Besides this, the sample also includes the “rapid communication” Physical Review Letters with the declared policy that topics have to be “of broad interest” -- clearly not representative for physics by large, if you excuse the sarcasm.

But to understand what the authors mean with “hot”, let us look at what they have done. They quantify the physicists’ ‘tracing’ of hotness by the probability that the subject of a new paper depends on the number of papers already published on the topic. Topics are identified by the PACS number of the paper (a paper can thus belong to several fields). If new papers are not evenly distributed over existing topics, but those topics with many publications already are more likely to attract new ones than random chance would suggest, this is known as preferential attachment. It’s more commonly known as the “rich get richer effect” and can be quantified by fitting a power-law to the distribution.

The authors find that the physics papers in their sample do show preferential attachment, ie who has will be given. The effect is not as pronounced as for some social networks (eg Flickr) where similar studies have been done, but it clearly exists. They have further looked at the scaling in subsamples broken down by the country of origin of the first author and done the same analysis separately for a selection of four countries: Japan, China, Germany and the USA. They find that the preferential attachment is the strongest for China, followed by Japan, Germany, USA. Yes, that’s right. According to this study, Americans are less likely to follow “hot” topics than Germans.

In the introduction of the paper the authors remark “It is believed among many scientists that there are many more Chinese scientists that are followers than original thinkers compared with many other countries.” I find this an interesting statement for a scientific paper, seeing that it’s little more than spelling out a perceived stereotype. Though they may be forgiven their bleak view of Chinese scientists since, for all I can tell, the authors are all Chinese, or are at least working in China. They interpret the results of their study as confirming this stereotype.

It should be mentioned that the sample which the authors analyzed also contains comments, replies and errata that I’d have thought should be mostly evenly distributed over topics. I would guess if these were taken out the sample, the overall effect would increase somewhat.

But is this preferential attachment a sign that physicists follow “hot” topics?

What this analysis actually shows is that Physical Review preferably publishes papers on topics that already have a literature base. I wouldn’t call that tracing of “hotness”, I’d call it conservative. If you wanted to quantify how eager physicists are to jump on ‘hot’ topics, you’d have to measure how likely new papers are to be in rapidly growing fields, as opposed to fields with many publications already. And to add my own perceived stereotype, I’d be very surprised if you’d find the Germans jump faster than the Americans.

In summary, this study isn’t uninteresting but the interpretation of the data is highly misleading.