## Subject: Free duck with your oligos

A sample of today's biospam:

Mark, get your ducks in a row with Operon…
FREE Operon Op-Animal with ANY oligo order!

Operon and MWG are joining forces to create a new global leader in the oligo market. Operon has created the world's most advanced factory for fully automated solid-phase synthesis of oligonucleotide probes and primers.

And they'll send you ugly plush animals too.  Shockingly, these are apparently popular enough that people try to to merge orders so as to reach the 150€ minimum to get one.

I wonder which of my internet actions put me on this mailing list? I suspect that my mistake was subscribing to Nature and/or Science. Apparently it's not enough that they mainly publish biomedical articles, so as to appeal to the biomedical research equipment and services companies who dominate their ads; perhaps they make extra money by selling their subscriber list to the same companies?

I don't know that either of these publications is guilty, and I apologize for the suggestion if they're innocent — after all, it could be co-authoring a paper in BMC Bioinformatics, or attending some bioinformatics conference, that signed me up as a biospam target.  But anyhow, this underlines the fact that at the top of the scientific-industrial food chain, there is a third funding model: not only "author pays" vs. "reader pays", but also "spammer pays". Well, to be fair, "advertiser pays". An obvious consequence is that this warps the editorial policies and the journal content in the direction of the audience that the spammers advertisers want.

And of course, in the end this is largely a form of "government pays", since much of the money for the expensive research equipment and services comes from government research grants.

Unfortunately, there are many disciplines — from sociology to mathematics — whose practitioners' grant money is rarely spent on expensive products and services offered by companies who advertise in Nature. Imagine if PCR stood for "phonological chain reaction", a method that takes a few audio clips and uses specialized and expensive equipment to produce an unlimited amount of babble. Free duck with your phonemes!

Oh wait, there is such a method — but we're unlikely to see anyone advertising "the world's most advanced factory for child language acquisition"…

1. ### mgh said,

June 14, 2008 @ 9:13 am

"Apparently it's not enough that [Nature and Science] mainly publish biomedical articles, so as to appeal to the biomedical research equipment and services companies who dominate their ads […]"

Usually, when someone else writes a sentence like that, I count on you to write a witty post where you test the hypothesis, for example by calculating the ratio of biomedical articles to (for example) geoscience articles in the relevant journals and comparing it to the relative sizes of the biomedical and geoscience research communities at a sampling of major universities, repeating that for a few other fields (or national funding levels), and giving us a nice chart of the relative representation of each field along with the relevant p-values.

I'm not saying your hypothesis is wrong (though to me it seems far-fetched), and I complain about the publishing system as much as the next person — but, as when accusing NBA coaches of collusion, I think it's a serious charge and it's the accuser's responsibility to provide some evidence!

2. ### Sili said,

June 14, 2008 @ 9:48 am

That's not a plush animal. These are plushanimals!

</spam>

Yes, yes, I know. I'm in breach of the Maxim of Relevance, but who doesn't want to look at big, adorable stuffed animals?

3. ### Mark Liberman said,

June 14, 2008 @ 11:11 am

@MGH: it didn't occur to me that there was anything in the least controversial about the observation that Nature and Science "mainly publish biomedical articles". I thought this was just common-sense observation of the facts of the world.

But here are some numbers. Picking the last four issues of Nature (v. 453, nos. 7193-7196) off of my living-room coffee table, I find that out of 13 published articles, 10 are biomedical, 2 are about cosmology, and 1 is about climate change. Out of 54 published "letters" (i.e. short articles), 32 are about biological or medical topics. The second-largest topic (10 letters) seems to be solid-state physics and related areas of chemistry (whose practitioners also need to buy expensive equipment). The third commonest topic is climate change (or climate history), where I presume that the motivation is general interest rather than advertising potential. There was one letter in what might be called sociology (reporting a study of "the trajectory of 100,000 anonymized mobile phone users whose position is tracked for a six-month period".) There was nothing, for example, in mathematical physics or high-energy physics; nothing in psychology (outside of biochemical or imaging-based neuroscience).

Looking at the ads in one issue, I find that essentially all of them — and there are dozens — are for equipment or services for biomedical research, along with a few for (mostly biomedical) journals and books, and a couple advertising geographical regions as a good place to live or to site research facilities. I admit that it's less obvious what direction the causal arrow should point, in relating the historical development of the scientific and the advertising content of such journals. However, I think that it goes without saying that most of the advertising revenue now comes from companies selling devices and services for biomedical research, and that this gives the editors a substantial motivation to publish scientific material that appeals to the audience that their advertisers want.

There's nothing wrong with this — it's one of the rational models for the economics of journal publication — but we shouldn't pretend that this situation is something other than what it is.

4. ### mgh said,

June 14, 2008 @ 11:41 am

Mark: To my mind, the null hypothesis is that the ratio of biomedical to non-biomedical articles in these journals is proportional to the ratio of biomedical to non-biomedical research ongoing today. It seems to me that the standard you hold other writers to would require them to test this null hypothesis before invoking collusion between the scientific editors and the advertising (or circulation) department.

In case the above remains unclear: I was not doubting whether Science and Nature "mainly publish biomedical articles" but whether that is out of line with what you'd expect from an unbiased sampling of today's research. (Another way to indirectly test this idea would be to ask whether the journal's rejection rates are lower in the biomedical sciences. A third way to test it would be to compare articles in advertising-supported journals to ones in non-advertising-supported journals, if such things exist.)

I'd emphasize that the charge that the scientific editors are colluding with the finance-oriented departments is a serious one (particularly when leveled at Science, which is a AAAS publication nominally run as a non-profit enterprise by and for scientists).

Does that make sense?

5. ### Mark Liberman said,

June 14, 2008 @ 12:17 pm

@mgh: …the null hypothesis is that the ratio of biomedical to non-biomedical articles in these journals is proportional to the ratio of biomedical to non-biomedical research ongoing today.

I don't know how to take an unbiased sampling of today's research. But however you might do this, I tend to doubt that it would yield a 42-to-1 ratio of biomedical research to social science research, for example; or that computer science research should be entirely unrepresented in 67 IID draws from the hypothetical Research Urn. You could argue that such publication statistics reflect (the editors' opinion) of the relative scientific merit of the work in these fields, but not that they represents the relative number of researchers or the relative overall volume of publications.

As for the relationship to advertising, conscious "collusion" is not required, any more than biological evolution involves anyone's conscious intent. All that is required is a simple and obvious application of economic (or Darwinian) reasoning: and when you see people and organizations acting in a way that is exactly matched to their material interests, it's logical to note the relationship.

I agree that the ethical issues (to the extent that there are any here) are different for the (for-profit) Nature Publishing Group and for the American Association for the Advancement of Science. But if there is any ethical issue in the latter case, surely it mainly falls on AAAS members and fellows, of which I'm one. And it seems to me that Science is pretty well adapted to its ecological niche, and there is little or no reason to interfere, especially given that the barriers to new publication opportunities are so low.

But again, there's no reason not to call a spade a spade, even though I have no intention of trying to reforge it into some other kind of tool.

6. ### Arnold Zwicky said,

June 14, 2008 @ 12:19 pm

To mgh: some things are so clear that it's a waste of time to demonstrate them. But here's a quick comparison of the 30 May 2008 issue of Science and the May issue of Scientific American, looking at ads rather than articles. (I started to look at the 24-30 May issue of New Scientist, but it was so strikingly similar to Scientific American that I didn't think the exercise was worth my time.)

The first point is that most of the ads in Science are not ordinary ads, but announcements of prizes and conferences and (predominantly) ads for services and products supplied by AAAS and/or Science magazine. There were 15 of these in the issue I looked at. The primary advertiser in Science is AAAS/Science.

There were 10 ads (8 of them full-page, so very noticeable) for biology/biomedical products and equipment, and exactly one other ad, for the bibliographic software Reference Manager.

Scientific American is utterly different: 5 biology/biomedical ads and 34 for other stuff, including: cars, cruises and tours, Rolex watches, the Visa card, a singles group for scientists, diamonds, books, the Rosetta Stone language learning software, and Touch of Gray hair treatment.

Counting articles would be a serious task, since both publications have a number of different categories for items, so that it's not clear what counts as an "article", and since many issues have themes, so that you'd have to look at a considerable number of issues to correct for this.

But here's a pilot study, using one non-thematic issue of Science (11 January 2008, chosen at random) and sticking to the articles labeled Reports (which are technical articles intended for specialists). The reports are categorized. I get an even split (7-7) between biolab stuff — lab biology/medicine/biochemistry — and everything else:

1. one Medicine, two Immunology, one Molecular Biology, one Biochemistry, one Cell Biology, and one categorized as Chemistry (but about the chemistry of RNA and DNA).

2. one each of Astronomy, Physics, Chemistry, Geophysics, Geoclimate, Paleoclimate, Ecology, Psychology.

(Note: Evolution is a category on its own, which i'd put on list 2. Some other categories for list 2: Materials Science, Climate Change, Planetary Science. Applied Physics, Atmospheric Science, Paleontology, Anthropology. Some other categories for list 1: Microbiology, Developmental Biology, Biophysics, Neuroscience, Cell Signaling, Structural Biology, Botany (for reports on lab studies). This world of categories isn't easy to negotiate.)

In any case, the list 1 reports don't literally predominate in the issue I looked at, but they're close, and they beat any comparable cluster of categories, and they're certainly prominent.

7. ### mgh said,

June 14, 2008 @ 12:50 pm

A place to start might be an NSF survey of PhDs awarded, for example
http://www.nsf.gov/statistics/seind06/c2/fig02-16.htm
(see link to "Source Data" at the bottom of that page)

According to their data, 8,250 PhDs were awarded in the medical, biological, and agricultural sciences in 2003, compared to 870 PhDs in computer science and 990 in mathematics. Obviously, one needs to take into account the relative publishing culture in each of these fields (my understanding is that, in computer science, having an abstract accepted at certain meetings is equivalent prestige-wise to what, in biology, is associated with a Science publication) and many other factors, but the publication numbers don't seem too out of whack to me here.

On the other hand, your intuition that the social and behavioral sciences are especially underrepresented is borne out: 7,410 PhDs in these fields were awarded in 2003, comparable in number to the biological sciences, while your survey above found only 1 paper you categorized as sociological (though as I recall it did get the cover :) ).

As to whether the editors actively bias themselves towards articles relevant to their advertiser base, I think it's an over-the-top charge. But then, I'm a trusting soul who likes to think the best of people. It's your blog, your opinion, and my subscription is free! Thanks for the discussion.

If you wanted to consider extending the quantitative side of this debate into a "meta" sort of commentary for Science, asking "why are social sciences underrepresented in the top journals," I think it would be a worthwhile contribution.

8. ### mgh said,

June 14, 2008 @ 1:02 pm

Arnold: again, to be clear, I don't doubt either that
1) most articles in the top journals are biomedical
or
2) most advertising in the top journals is biomedical.

I doubt that (2) causes (1).

Recall (for example) that the annual NIH budget is $30 billion while the annual NSF budget is$6 billion. It shouldn't be surprising that advertisers follow the money. The question is, whether the scientific editors at these journals follow the money — which, if true, is very disappointing and should change.

How would one go about such an audit? It seems to me that one would want to know the number of articles submitted to Science in each field, the number sent for peer review, the number that finally appear in print, the readership in each field, and the advertising revenue in each field, for starters.

9. ### Garrett Wollman said,

June 14, 2008 @ 3:36 pm

@mgh: Of course, the notion of what constitutes a "top journal" in a particular field is a self-reinforcing one: if Nature doesn't publish computer science articles, it's not a venue computer scientists are likely to submit to. The primary venues for computer scientists are, as noted, conferences, particlarly ASPLOS, PLDI, OSDI, HotOS, SIGGRAPH, NSDI, FAST, Usenix ATC, Crypto, and so on. Most of these conferences are sponsored by the IEEE, the ACM, SIAM, or Usenix (and often by more than one of them), and the ACM and the IEEE publish the principal peer-reviewed journals of the field. I'm sure that linguistics has its own preferred publication venues as well.

Nature and Science are the most visible outposts of the scientific publishing culture, but it's hardly surprising if they don't publish anything at all from some disiplines merely as a result of submission bias.

10. ### Rosie Redfield said,

June 14, 2008 @ 4:49 pm

Arnold: One of the most important lessons beginning scholars learn is things are rarely so clear that it's a waste of time to test them in a well-controlled study.

11. ### Arnold Zwicky said,

June 14, 2008 @ 6:12 pm

To Rosie Redfield. saying "One of the most important lessons beginning scholars learn is things are rarely so clear that it's a waste of time to test them in a well-controlled study."

As a general principle, of course.

You don't need to teach your grandfather (or, perhaps, great-grandfather) to suck eggs. But if *everything* has to be established, from the beginning, all the time, without respect for experience, by empirical data, we are all in very bad trouble. We're frozen.

It's a balancing act. But to move at all, we have to take some assumptions and perceptions for granted. Someone can come along with evidence to the contrary, and then we have to look at that. But just questioning the clear trend of things is only obstructive. (Yes, I know, I write a lot about how people's perceptions of what's clear are often mistaken, because of their (mistaken, and sometimes strongly held) beliefs about the way the world works. But that can't throw into dispute every perception about everything.)

Pretty much nobody is denying that the articles and ads in Nature and Science are heavily tilted towards lab bio stuff, and some of us have provided (very preliminary) evidence that this is so. (The discussion is about what this means.) I am frankly embarrassed at having wasted time on this; it's a lot like having to defend the claim that "while" can be used as a temporal subordinator ("While they were playing, the sun went down"), or that most written English sentences have subject-predicate structure, etc.