Andre Geim - he of levitating frogs and Nobel prize in chemistry for graphene, had a paper with his hamster as a co-author:
doi:10.1016/S0921-4526(00)00753-5
I remember as an undergrad using one of my professor’s books as a source in a philosophy essay. The book had been written some years before, and he’d modified his position since then. I’d had discussions with him about his new views which I’d referenced in the essay, but it was a nightmare trying to reference it. I think I ended up saying “as noted in recent coversations with the author” or something, so no idea whether that would pass muster for publication (rather than an internal university essay).
The standard way of citing something like that is as a “personal communication,” e.g. (John Smith, pers. comm.). Some journals these days are requiring that someone cited that way confirms that the information is correct.
Journal Editor: “robert_columbia, John. P. Bellows cited a personal communication from you in his proposed article, ‘Behavior of the tsetse fly when faced with quadratic hypotenuses in a hypowater bath’. Could you please confirm that the citation is accurate?”
robert_columbia: “Please add this number to your ‘do not call’ list.”
Journal Editor: “Sir, do you realize that by failing to cooperate, you may be jeopardizing the career of John P. Bellows?”
robert_columbia: “I don’t care. He violated my little sister, he can go to hell for all I care.”
Sure, that could happen (though you should realize you’re not helping your career at all if you’re a scientist). Even with the best of intentions on your part, you could also be unable to confirm the communication because you’re in the hospital with sleeping sickness or off in Africa doing field research or something.
But it’s OK that citing personal communication is harder and riskier than citing published material, because we don’t want papers to rely on personal communications. We want them to reply on published verifiable information, and only include cites to personal communication as a last resort. After all, if the journal editor can’t confirm the information, then neither could anyone reading the paper, and that’s bad science.
To what extent should $10,000 journal subscriptions and $40 articles be considered “inaccessible”?
As things stand, they are inaccessible to most people, but reasonably accessible to faculty and students at a decent research university in a first world country (if only by inter-library loan, and that sometimes for a small fee). The latter group probably form the considerable majority of those who would be interested, but they are not everyone interested, certainly, and the rest are effectively cut out.