for background on PeerJ see here https://www.loomio.org/d/VIgtdbXP/call-for-questions-for-peerj

I spoke to PeerJ founders Pete Binfield (who was formerly the executive editor of PLoS ONE) and Jason Hoyt.

PeerJ is a mega-journal in the style of PLoS ONE or Scientific Reports. For computer science, they created essentially a new journal, PeerJ Computer Science, running on the same infrastructure and tightly linked to PeerJ. They are not interested in doing this for perception/vision however. The numbers of potential authors are too low for them to have any confidence it would yield a vibrant community. The pool of potential authors in computer science is >>100,000, much bigger than the few thousand (?) in perception science.

But they are interested in creating sub-sections of their journal that are sort of like sub-journals but not as autonomous. For PeerJ, these provide several advantages over an undifferentiated mega-journal. They mentioned two such options:

Collections

So far collections have been created for papers on a very specific topic, or for those associated with a conference. They can be normal, peer-reviewed manuscripts accepted by PeerJ, or PeerJ preprints, including posters. An example is "The biology of the Hawaiian Archipelago" collection, an example of one for a conference is https://peerj.com/collections/15-msw15/ . The way this works currently (they seem open to potential modifications) is that an editor who proposes the collection decides which already-accepted articles are part of the collection and which aren't. This could be a good way for the community to dip our toe into the waters at PeerJ.

For example, ECVP might decide they'd like to highlight the award-winning posters by creating a PeerJ collection of them. The authors of said posters, should they like the idea of being included, would submit their posters to PeerJ Preprints. A person from ECVP would then approve them for inclusion in the collection. (I don't know whether ECVP is giving poster awards this year, and perhaps if they do only give a few, not worth a collection, but one idea is to include an entire "longlist" of posters that were in the running for the award). An alternative is to create a Collection that is essentially a special issue, if we can think of a topic that would attract good authors.

This is a purely informational post, I won't advocate here for my preference- a PeerJ perception subject page :)

Subject pages

A subject page is more like a journal within a journal. A subject page points to all the articles accepted into PeerJ on a particular topic. There is one for Psychiatry and Psychology, which is mostly automated and has a "Latest articles and preprints" section and feed, also "Widely read", etc., and "Editors' picks" which I believe by default is those articles highly rated by the accepting editors. Along the right is a list of "Editors and advisors".
There is no editorial content in the sense of stuff (editorials) written by editors but PeerJ is open to that, I think they said they're planning on adding functionality for that. The idea is to create a community feel, and I think we / the editors of Perception/vision within PeerJ would want to be able to write editorials.
Notice the ways that a PeerJ Perception or Vision subject page would differ from a separate journal

• Collections and Subject Pages don’t assign their own ISSN nor citation. The citation for each article would be to PeerJ (the journal) or PeerJ PrePrints
• Same editorial standards as the rest of PeerJ, which is essentially just like PLoS ONE, importance not a criterion, just good methodology. It is quite subjective ultimately what counts as good methodology though, and basically the editors in the area decide that. However there currently is no facility for a chief editor of a subject, so there is no built-in across-editor organization.

Other issues

• Institutional funding to reduce the (already-low) author fees? It's possible. #1, institutional memberships exist, e.g. Jon Peirce said his uni is a member, making publishing at PeerJ free. Also, bulk buying already exists for handing out to authors of your choice.
• How do we know they won't sell out to Elsevier? They said they unfortunately can't guarantee anything. One of the co-founders, Jason Hoyt, came from Mendeley and from his post at the time you can see he was very disappointed (somewhat guarded, he was talking about his colleagues) in them selling out to Elsevier http://enjoythedisruption.com/post/47527556151/my-thoughts-on-mendeleyelsevier-why-i-left-to
• They don't publish traditional review articles where an author surveys the field and writes their opinion. They do publish however things that look like meta-analyses or that result in plots of evidence mined from the literature, e.g. https://peerj.com/preprints/921/
• They have nice support for embedded movies which they host themselves, and create both small download and large download versions.