Middlemen

In both art and science, the products of (small groups of) individuals are disseminated to the world by other companies. In the world of art, these companies are the galleries representing artists. In the world of science, they are the publishers and their journals.
In both these situations there is a clear distinction between those who produce the goods and those who distribute them to a wider audience. The presence of such middlemen is common in many industries, but an uncommon aspect found in both art and science is that the financial benefits to the intermediary are far greater than those of the producer. Scientific publishing is now a multi-billion euro industry and the largest of the art galleries have turnovers in the range of tens of millions of euros.
The curious similarities between the two fields are the result of imperfect information on the consumer side, combined with some leftovers from an older world where the financial risks were differently distributed and legally arranged. 

For both scientific publishing and art galleries the most valuable asset is the firm's reputation.
For example, the price of an artwork is linked quite directly to the standing of the gallery it is shown in. Similarly, a scientific discovery is generally considered more impactful if it's published in a journal of significance. It's therefore imperative for both galleries and scientific journals to become, and remain, reputable. It's also easy to see that for both fields there is simultaneously no inherent and necessary connection between the quality of the work and the social standing of the middleman. The intermediary does not change any intrinsic property of the final product. That the perceived quality of the intermediary is nevertheless seen as a useful indication of the quality of the good is due to a characteristic that economists call imperfect information.
In both art and science, there is no information about the quality of a good that is both reliable and readily available. The causes of this imperfect information are different in each field, but over the course of the last century they have led to a similar outcome where the intermediaries have a disproportionate influence on both the kind of goods that get produced as well as which consumers have access to it.

Any consumer needs information about a good in order to make a decision about what is worth spending their money on. They can either have full access to all necessary information, which is called perfect information, or limited access to one or more characteristics of the good, which is called imperfect information.
In both art and science, information about the quality of a good is difficult to ascertain for a large number of interested buyers. Quality in the arts is next to impossible to quantify and subject to changing cultural perceptions. And while scientific merit can be checked in principle, this requires an impossibly large amount of time, money and other resources, so in reality it is unfeasible for any one party to make an objective judgement based on their own experimental knowledge about the quality of all articles published in all journals.
Hence, for both art and science, there is a lot of effort that goes into convincing a potential customer of the value of the good that is being sold. As the goods themselves don't provide accurate clues to their genuine value, this is done through less direct means that convey a perception of longevity. Such (signs of) longevity can ostensibly only be reached by consistently providing quality goods.

Galleries and artists today try to provide credible signals of value by demonstrating a long-term commitment to each other. Until the early 20th century, this meant that dealers were directly buying (nearly) all of an artists output, thereby putting their money where their mouth is. If nothing else, this would at least demonstrate to a potential client that the dealer has a strong belief in the artist. And the dealer is only able to aquire those works today if they have made sound financial decisions in the past. These days the commitment is less strong, expressed by 'representation' of the artist by the gallery. The financial capabilities of the gallery are generally demonstrated through large, and mostly empty, spaces in expensive buildings in desirable locations, as well as participation in ludicrously priced art fairs.
I've written elsewhere on this blog how this shift is likely partially caused by changes in anti-trust legislation in the Western world in the first half of the 20^th century, so I won't delve any further into this subject here.

The publishing of scientific works likewise underwent significant changes in the last two centuries.
A scientist is usually employed by a university or some other institution, and when they've made a discovery, they write up what they've done to try and make it known to others what they have discovered. It is of course difficult to reach a broad audience, even in the age of the internet, so this is one thing a publisher can help you with. A publisher possesses ways to reach an audience that any single scientist doesn't have. A publisher also has access to infrastructure. Although these days more and more of scientific publishing is done digitally, physical printing and distribution of materials has historically been a venture with large upfront costs, combined with specialised knowledge and equipment. These upfront costs carry significant financial risks, which can only be borne by a large company that is able to spread such risk over multiple ventures. 

It's also a well known fact that most scientific literature has a very small readership. Current estimates on the audience size of the average journal article range from single to triple digit numbers.
Yet at the same time, there is a great number of scientific articles that are published every day. With such fragmented readership, there is little possibility for scientific texts to gain widespread attention in the same way a newspaper article or a viral video might. Therefore, to a broad audience it is virtually unknown what the value is of any given article relative to all the other articles that are available.
As already stated, some of this uncertainty is remedied through the reputation of the journal the article is published in. This reputation is mostly based on the reception of the works that were published by the journal in the past, as well as the academic standing of its current editor(s). There have been attempts at quantifying this reputation by metrics like a journal's impact factor, which essentially measures how often articles from a journal have been cited by other scientists. But as Goodheart's law states, any measure that becomes a target seizes to be a good measure, so such undertakings merely repackage the problem instead of solving it. 

Both industries thus have a small customer base and these customers ought to be sceptical of the goods they provide and the high prices they ask for them.
So how do these middlemen leverage their position to create profits for themselves?
In the arts it is a simple question of gallerists charging very high commissions for their work, so that a handful of sales can provide an adequate amount of turnover, especially when their risk is spread over a number of artists.
In scientific publishing, exorbitant profit margins only arrived around the turn of the millennium, and to see why this is the case requires a short history lesson on copyright law, and in particular how such laws were implemented in the United States of America.

The foundation of today's copyright legislation was laid at the Berne Convention in 1886. This type of copyright is based on an idea of author's rights, where the creator of intellectual property is also automatically the owner of intangible rights relating to that work. These reproduction rights could then be licensed to a third party, such as a publisher. This can happen in different ways, but it must be noted that a perpetual exclusive license to reproduce the work is an option, even when the author retains the copyright in such a case.

This is in contrast with the common law idea of copyright, which is much more focussed on the economic right to publish and distribute. The United States, which legal system is based on common law, was thus late in incorporating the principles of the Berne Convention. In the early 20th century, copyright for individuals did not exist, but a publisher could register the publication of a work at the copyright office to obtain its copyrights.
It's a bit of an oversimplification, but it wasn't until the Copyright Act of 1976 that the intellectual property laws of the United States became more closely aligned with those of most other countries.

This change has quite directly lead scientific publishers to mandate their authors to sign over their copyrights to the publisher, instead of licensing their papers. At best this can be seen as a good-natured attempt to retain the best publishing standards possible, but it's much more likely that this decision was aimed at retaining control over the substantial captial the publishers had ammassed up until that time.
For example, in the 1966 edition of the Handbook for Authors of Papers in the Journals of the American Chemical Society, the section on 'Liability and Copying Rights' is only half a page long and simply states 'The Society owns the copyright for any paper it publishes'. This was true under the federal copyright laws of the USA at the time, which required registration at the copyright office.
Interestingly, the section on 'Liability and Copying Rights' of the 1978 edition of the Handbook for Authors of Papers in the Journals of the American Chemical Society was nearly twice as long as the previous edition. It now contained the following phrase: 'Under the terms of the Federal copyright law, effective January 1, 1978, scientific publishers who wish to obtain copyright ownership of papers in their journals are required specifically to obtain such ownership from the author of each paper. Since it is necessary for the widest possible dissemination of scientific knowledge that the society own the copyright, authors are required to transfer copyright ownership before publication of their manuscript.'

This last sentence is simply not true. A perpetual, even exclusive and non-restricted, license to publish poses no practical objections. However, such a license would still leave the ultimate ownership in the hands of the author, so that the publisher could not license the work out to third parties. Transfer of copyright ownership thus is an issue of control of the work beyond the any direct publishing efforts in their own journals.

However, it might have been vital for publishing companies to protect those interests. In the 1970's, publishing was still a complex and costly business, with large upfront costs and little or no guarantee that anybody would be interested in the final product. The publishing industry had a high risk of failure and the small number of scientific publishers that have survived, only survived because they originally published books that turned out to be of particular significance and relevance to other scientists. Unlike most of their publications, these tomes had several reprints and made a healthy profit, which could offset the cost of the many failures.
It is, however, impossible to predict which publications become a hit, and if the publisher didn't own the copyright, such a reprint would have most likely have to be renegotiated with the author. This author by then has of course seen how well their book is selling and so they'll likely want a bigger cut for themselves, or could even let the reprint be done by a different publisher altogether. This is therefore debilitatingly risky to a publisher during that time, and so the transfer of copyright ownership may have been a reasonable request in the 1970's. 

This all changes towards the 2000's and the advent of widespread internet access. Through more than a century of publishing and consolidation in the industry, a handful of scientific publishers are in possession of enormous archives and because of their insistence on copyright ownernship, they have full control over them. Through digitisation these materials are now also easily searchable and through the internet they can be distributed at negligible cost to the publishers. In other words, the publishers' material capital is now more valuable than ever, while their operating costs have fallen dramatically.
As a result their profits have risen to extraordinary heights. To illustrate this fact, two of the top ten highest paid CEO's in the Netherlands are scientific publishers. 

In summary, the presence of middlemen is necessary in both art and science to create credible signals about the quality of goods. And in both these markets, the middlemen have understood the necessity of their presence and found ways to leverage that power into great profits by essentially exploiting the weak negotiating position of their suppliers and in some cases those of their clients.
Such predatory practices are much lamented in both industries, yet I'm unaware of any proposed solution that could remedy the problem. Many of such initiatives are focused on the (financial) inequality of the artist/scientist and gallery/publisher, but I believe a solution can only be found in making information about the quality of goods readily available to end users.

As a final remark, it must be noted that book publishing in the arts is a market that functions remarkably well, considering the difficulties that exist in scientific publishing and the sale of artworks. In art publishing, there is a healthy market of buyers and sellers, while risks and profits are usually shared in reasonable terms between the artists and the publishers.
The reason for this is simply that an art book can quite literally be judged by its cover. When selecting which art book to buy, an interested buyer usually can find the books that appeal to them by considering the design of their covers. Art books also retain their value rather well, so that even if a mistake is made, a buyer can still resell the book at only a minimal loss. Therefore information on the quality of a good is widely available, while the cost of misinformation is marginal.
This is the exact opposite of scientific publishing and the market for artworks, where credible information is hard-won and the costs of getting it wrong can be extremely high.