Do journals in general have any kind of policy regarding papers submitted by someone without a research affiliation?

Some journals implement a double-blind reviewing process, meaning that the reviewers are not aware that the authors are from academia or not, and only the scientific content is judged. That being said, it's worth mentioning that it would be hard for someone without a proper "paper-writing" training (such as the one one can acquire in academia), to produce a paper that would be accepted by reviewers. Some general structure is expected, such as related Works, critical discussion, rigorous methodology, and I would say that without that, it would be hard to get the paper published (I have myself rejected papers from graduate students, not because the idea itself was bad, but because the structure and the presentation were not meeting the standards one could expect for a scientific publication).

EDIT: After reformulation of the question, assuming that the quality of the paper makes it indistinguishable from any other paper, then, no, as far as I know, there is no general policy regarding the official affiliation of the author(s). For instance, in Computer Science, it's not rare to see papers published by people working in a "normal" company (i.e. not a research company), typically on some concrete problems/solutions they have found. Some people even keep publishing after starting their own startup, and therefore the affiliation is something like "MyCompanyWeb2.0".

As far as I know, in just about every field it would be considered wrong to reject a paper merely because the author doesn't have a university affiliation.

However, as other answers have pointed out, a paper from a non-academic may have difficulty complying with the usual norms of the field for writing, and might be summarily rejected (or sent to a referee but quickly turned down after a cursory reading) on those ground. Mathematics probably gets more amateur submissions than most fields, many of them downright crankery---although now that I think of it, it's been several months now since someone e-mailed me a one-page proof of Fermat's Last Theorem---and coping with them is a problem, in part because there is a culture that says formal journal submissions need an actual reason to be rejected, just in case this is really the one time a genius from outside academia has solved a problem. But the converse is that as soon as such a paper does fail to comply with ordinary norms of writing and argument, there's grounds to reject it.

(By the way, your example is an interesting one because a study on memory presumably involves research on humans, or at least animals, which might have compliance issues---a carpenter has presumably not gone through the usual process of having research with human subjects checked by a board for ethical compliance, and I'm not sure journals would publish research which requires ethics review but didn't get it.)

I left academia about 27 years ago after completing my doctorate and a couple of years of post-doc. Since then I have published about 6 papers on physics and mathematics in peer-reviewed journals so it is certainly possible to publish without any affiliation.

However, this week for the first time I experienced the rejection of a paper purely because of the lack of affiliation, so I can confirm that some journals are now rejecting manuscripts purely on this basis. The paper I submitted was arXiv:1401.8217 which reports my progress on Lebesgue's Universal Covering Problem including a new upper bound. This work is not going to make any seismic waves in the world of mathematics but it is a well known hundred year old problem and previous improvements on the upper bound have been published and well cited.

I submitted to the Hindawi journal "ISRN Geometry" because it is open access and currently has no article processing charge. These are useful conditions for someone with no funding or easy access to subscription journals. It was also convenient that they do not ask for TeX layout and will do the formatting to their style themselves. I had written in Word and was glad not to have to reformat.

After about two weeks I received a message from someone at the editorial office to say that I just had to submit a new manuscript including my academic affiliation for the review process to begin. I was given two days to do this. I replied quickly to say that I had no such affiliation. Two hours later I received a final rejection notice "I regret to inform you that it was found unsuitable for publication in Geometry." No more details were given. The paper had not been sent to a reviewer and it was clear that the only reason for the rejection could be the lack of the requested affiliation. The person who signed the message does not have any academic record that I can find and is not one of the long list of editors for the journal.