Suppose a raging barbarian cannot reach any seen opponents on the battlefield on their current turn, but is attempting to maintain rage in the interim by attacking a hostile creature, according to the following:
Your rage lasts for 1 minute. It ends early if you are knocked unconscious or if your turn ends and you haven't attacked a hostile creature since your last turn or taken damage since then.
Suppose the player is making a good-faith attempt to maintain rage because they are still in combat, just unable to reach visible opponents within one round.
The barbarian declares an attack on a hostile creature they suspect is within their reach but which they cannot see.
If there was a successfully hidden opponent on the field, then RAW the barbarian would be permitted to attack it by guessing its location. Even if they were incorrect about the target's location, that would be sufficient to maintain their rage. We know that attacking near a present target is allowed, and the barbarian needs no surety that the creature is actually in the location guessed.
But if the successfully hidden opponent has actually left the field without the barbarian knowing, there is not a creature to attack.
Is the barbarian permitted to make an attack against an opponent that is not present?
I am trying to understand whether a creature actually needs to be present for the barbarian to attack, and if so, why its presence matters, or matters more than the barbarian's intent.
To me, either ruling, yes or no, has unfortunate implications.
If attempting to attack an opponent that is not on the field ends the barbarian's rage, that allows rage to be used as an 'enemy presence detector', which seems to go against the spirit of "If the target isn't in the location you targeted, you automatically miss, but the DM typically just says that the attack missed, not whether you guessed the target's location correctly."
But if the barbarian is allowed to maintain rage by attacking an opponent that is not actually there based on the plausible belief that an opponent is present, then what prevents the barbarian from postulating an opponent who could be there? For example, the barbarian invokes an NPC that has successfully hidden against the party before. The barbarian's belief that said NPC is present and Hidden can then become a source of conflict between the player and the DM, in trying to decide what is a reasonably imagined unseen opponent.
I am not asking about a bad faith attempt by a player to invent opponents that don't exist.
Somewhat related: A barbarian's belief that they are attacking an opponent is apparently not sufficient to maintain rage if what they are attacking is an illusion. So attacking a not-creature that is there is not enough to maintain rage, but is it enough to attack an actual creature that is not there?
Best Answer
(This started out as a comment on another answer, but was getting too long to be a comment, and threatened to turn the comments into an open discussion.)
This Is Exactly What GMs Are For
This is also why "Rules As Written" is not always the end of the discussion, and why TTRPGs are not CRPGs.
The rules are clearly not intended to allow barbarians an easy loophole to extend their rage to a full minute each time. But the rules are also clearly not intended to turn the barbarian into an illusion detector or a departed enemy detector.
Those are loopholes, and two of the three are metagaming loopholes intended to extract information from the interactions of the mechanics to the player in ways that the characters themselves shouldn't really be able to do. This is the sort of thing that would get lampooned in Order Of The Stick.
The obvious patch to the rule is that the barbarian has to make an attack against something he or she believes is or is controlled by a hostile creature. But that, of course, opens up the discussion of player agency, character agency, and the limits of the GM to impinge on those-- that patch, also, would not be intended to allow metagaming just by the player's fiat of the character's mental state.
Much ink has been spilled on that subject, and I try pretty hard to defer to the player in this regard, but that deference is not infinite, and it ends when I think there is an attempt to metagame the rules. And believe it or not, most experienced GMs have a pretty good sense after a while for when their players are trying to game the system.
Would I feel good about telling a player what is and is not reasonable for their character to believe? No, because it's a sign that the GM-player relationship is getting adversarial rather than cooperative.
Would I feel comfortable doing it? In certain circumstances, yes-- more comfortable than not doing it and perhaps enabling larger problems down the road. I would be more likely to open with a sharp, "Quit trying to game the system, please," rather than a flat fiat... but I'd be willing shove, if push came to it.
Someone has to make a judgment call, here, and the GM is the only person whose judgment call is final.