That’s a different relationship. It’s in edit release (not relationships) and shows up on the right under “Labels”, along with the catalog number. There are also a bunch of release-label relationships in the relationship editor, not just phonographic copyright.
I don’t think anyone is saying we shouldn’t link the releases to labels.
I’m pretty sure it’s a statement, that’s why we (for example) set both the begin and end year to 2017 for © or ℗2017. Not to mention representing the span is very difficult as it varies by country and changes fairly regularly.
Not really sure how that follows from it being a statement of copyright. Releases can state they claim copyright on the compilation (thus on the release) or on the individual recordings (thus on the recording) or, even, on both.
Saying “whatever! we don’t care! Multiple dates is ugly! Put it on the release” also doesn’t work. For a fair number of releases, different tracks have different ℗ dates and companies. Having a pile of ℗ dates and companies on a release without any way to tell what tracks they belong to is far less useful. So essentially then we’d be left with using the release annotation for any hope of accurate information—and that’s also far less useful, as its not standardized or easily machine parsable.
Take a look at https://musicbrainz.org/release/db77f67d-9e9b-4623-8ea1-baab173cbe49/cover-art — the booklet gives ℗ dates and companies for each track on 19 CDs and then a separate ℗ date and company for the compilation. By putting track ℗ on the recordings and the compilation ℗ on the release, that is essentially all represented in the MB schema (the only thing not represented is which of the dates is the digital remaster). Your proposal would have (quick guess) 40+ ℗ relationships on the release and convey significantly less information. It would also mean that the work that went into entering that information can’t be shared with other releases of the same recordings, even things like https://musicbrainz.org/release/2546da4f-9402-4ada-9426-3a79b8ed2e5e which share the same discs (down to disc ID and content, at least).
Copyright law considers anything that has some minimal level of creativity in it a new work, eligible for a new phonographic copyright. Each of those typically only has one date (though if its worked on by one person/company over the course of two years it may wind up with two years). But at MB, we consider all those different (from copyright’s perspective) works as one recording. So a recording will naturally end up with multiple ℗ dates, potentially from many labels. So even with first-date-only, you’d still get a bunch of dates/owners on the recording. (There are more reasons to get multiple, of course, that’s just one).
I don’t know if you’re being facetious, but in modern releases, copyright holders are nearly always different MB Label(s) than what we’d use as the release label. If you need details or annotated examples, please ask.
I’m pretty sure that no one is proposing that ℗/© statements that are clearly given specifically for a given Track/Recording should be related at the Release level. We were talking about «℗ 19XX Label» vs. «(This) Compilation ℗ 19XX Label» both on the release level (see @jesus2099’s post). @jesus2099 was suggesting that if the ℗ statement does not specifically mention “compilation” (or similar)" in the ℗ statement that it can be inferred that the ℗ applies to all the Tracks/Recordings on the Release instead of (or in addition to) the overall/general Release itself.
Yes, I was a bit with the release label one, but not with all the other release/label ARs. If the full details are put in, there will still probably be a relationship to the companies on the release (e.g., from the © statement).
Ok, I though @Kid_Devine was arguing that. If not, I’ve misunderstood what he/she was saying, and apologize for attacking a straw man.
Ok—let me see if everyone agrees. I’ve never played the the poll tool before, so… an excuse to play with it 
Obviously, some of these conflict (in particular, 3 vs. 4 and 5 vs. the rest). I’ve set the max to 3, as I think you can’t pick more than that without contradiction.
3 vs. 4 is intended to capture what the default should be—℗ defaults to recordings or to releases.
0 voters
I’m somewhere between #4 and #5 in the vote. I don’t like «if you have reason to think», that’s too vague. If you don’t know for sure, don’t add them to the Recordings. It’s better to play it safe (ie., add on Release level) than add them on the Recording level and later find out it doesn’t apply to (all) Recordings and we having had bad data in the meantime.
So I’d say something like:
If a release gives “℗year company” for the whole release, not explicitly saying which tracks it applies to, put it on the release. If you determine for sure that it applies to all tracks, you can put it on the recordings instead of or in addition to the release.
I have to agree with Freso here, also with the wording he proposes.
Necro-posting here, but having read through this, I’m left wondering, so I bring it up here, in light of recent reasonable pushback from @jesus2099 on some edits (e.g., https://musicbrainz.org/edit/101103701) I wish to resurrect this discussion & gain some clarity.
If I’ve missed this being in the style guidelines somewhere, please let me know, but I think I scoured thoroughly & am left wanting.
For me, I attach copyright & phonographic copyright to the release, not the individual recordings, unless a release specifically tells me otherwise, I view the copyrights as applying to the release not the individual recordings contained within. This is mentioned in the poll above as
“℗ relationships should only go on releases, not recordings (with the exception of standalone recordings.)”
which does appear to be less popular than
“If a release gives “℗year company” for the whole release, not explicitly saying which tracks it applies to, it should generally go on the recordings, unless there is some reason to believe it doesn’t apply to all of them.”
but it’s unclear to me if this poll result is final, or if further conversation happened elsewhere.
Thanks to @jesus2099 for their helpful critique of my edits & encouraging me to think about this 
I not read the thread, but usually I’ll put a (p) at only Release level. With only a few changes to that. Sometimes an album may include one or more tracks from a different source. If the credits then say “tracks 1-10 (p) blah blah, track 11 (p) someone else” then I’ll do all the (p) at Recording level instead.
I know some editors who always put all the (p) at Recording level all the time. I see the logic in that and would not delete. It can be interesting when you look at some Recordings as this can lead to three or four (p) credits from different sources.
Look at an Album like The Dark Side of the Moon. You’ll find four different (p) credits on the tracks. Caused by each time they reissue it: Pink Floyd - Money
The Gramophone Co. Ltd. (in 1973)
EMI Records Ltd (in 2003)
Pink Floyd Music Ltd. (in 2011)
Pink Floyd Music Ltd. (in 2016)
It makes sense to show those all on the recording as some people like to track where the money goes.
especially “when we all pull together as a team” 
I pretty much would always add (p) at the recording level unless I have a reason to believe it does not apply to all recordings (as per our usual relationship guideline “Prefer most specific level”).
I agree.
Furthermore I’d definitely never enter an edit to remove the relationship from recordings unless I have really good reason to believe that they are wrong.
That’s a logical way I fully agree to and enter it in my edits, though it could lead to multiple entries in recordings as in CD1 of https://musicbrainz.org/release/ba59bf45-1a7a-45b4-b9cd-3d1389b2c0f5.
Yes it certainly does, and very often.
But I don’t think it should be seen as a problem. It is rather a sum of informations.
Like it’s fine to have several ISRC on recordings and fine to have the same recording in several releases.
Sometimes I enter phonographic copyright at release level. For example, if there’s a ℗ 2020 on a reissue while there was no copyright on earlier releases (or only ℗ xxxx without company). I think, 2020 would be confusing for a 1970s recording.
If it is something like a reissue of The Dark Side of the Moon, then I can see the sense of adding the 2020 date to a recording. This is Pink Floyd nailing down a copyright date again. I think this is part of why they say they are “remastered” so they can claim “new recording, new copyright” and restart their 50 year clock.
If it is a compilation like NOW Hits then I think it would be wrong to add the (p) date at recording level. For something like a compilation usually the booklet is full of separate dates. And at that point I am then confused if they should be part of the Work or the Recording… so just skip it.
…and that’s in addition to the original copyright. 
I thought more about something like this → https://musicbrainz.org/release/36b49a1c-f0de-3033-87a0-22b77836e6d6
In this case it is already attached to the recording ¹, but the 1999 date comes from a reissue. The original albums had no phonographic copyright claims.
¹) and it seems to be wrong, should have been “Sony Music Entertainment Inc.” 
Someone somewhere will have a (p) claim on the original. May get filled in one day. I’d leave the 1999s in place as they are technically correct, even if not first. (You also look right on the wrong Sony being selected…)
I guess the money grabbers followed different rules in the 1960s and didn’t splash their cash claims so loudly over everything.
The " Pink Floyd - Money" example I posted earlier is the kinda thing I meant. The labels keep reissuing it, and each time a new company makes a claim on the cash. Always makes me laugh that there are so many Floyd tracks and albums about the greed of the industry… and they are now among the worst for the constant boxset cash-ins. 
Probably, but there was no ℗ notice in 1968. It was introduced in 1971. And many countries have very different legal protections for sound recordings (and for filing claims). Up until the 1980s it was completely unusual in Austria and Germany to print something like this on releases. But as far as I know, monetary claims were not forgotten in the 1960s either 
Of course I leave it, it’s totally correct …or it will be correct when my edits are applied.