Back in the 90s when I played 2E, before Third Edition was released, we played with perception as a seventh ability score. I’m not sure if this was inspired by something, or if it was a purely original house rule. We relied heavily on “roll equal to or less than” ability checks (even basing a custom system entirely around this mechanic at one point). While the traditional six ability scores cover most situations, it is not obvious which should be used, for example, to notice a sneaking monster. So, we added another 3d6 score for perception.
The WotC editions of D&D were designed similarly, and I’m sure many (most?) other games work the same way. It seems to intuitively make sense that being perceptive, being able to notice things, is a personal trait, and should vary from character to character. However, this doesn’t really work very well in the context of an adventuring party. For one thing, it replaces 1 roll with N rolls (or N lookups if you are playing with something like passive perception). Further, a well designed group game will not have many opportunities for one character to go off and do things without the participation of the others, so the utility of tying perception to the individual is limited.
There are two main places where perception is relevant to the game. The first is when a PC is trying to locate something that is concealed (either actively or passively), the second is when one group might not be aware of another group (like an ambush). The original game used a per-character system for the first and a group system (surprise) for the second. From The Underworld & Wilderness Adventure (page 9):
Surprise: A condition of surprise can only exist when one or both parties are unaware of the presence of the other. Such things as ESP’ing, light, and noise will negate surprise. If the possibility for surprise exists roll a six-sided die for each party concerned. A roll of 1 or 2 indicates the party is surprised. Distance is then 10-30 feet.
Surprise gives the advantage of a free movement segment, whether to flee, cast a spell or engage in combat. If monsters gain surprise they will either close the distance between themselves and the character(s) (unless they are intelligent and their prey is obviously too strong to attack) or attack.
There is precedent for using the surprise mechanic as a general stealth system. This way of doing things was preserved with minor variations through all the TSR editions of D&D. For example, in Second Edition, the check is 3 in 10 per side (2E PHB page 111); there are options that allow individual character attributes to influence the group roll though. I didn’t understand or use the original surprise system back then though, for whatever reason.
Perception as we used it (and as D&D 3E and 4E use it) is a highest common denominator system. You only need one person to notice a sneaking bugbear and then they can alert the other PCs. In contrast, movement rate is a lowest common denominator system. The group can only move as fast as it’s slowest member. Trap finding is also a highest common denominator system. If connected to player options, any kind of highest denominator system will structurally devolve into a skill tax when taken to its logical end. That is, an “optimized” group will have one character that is perceptive (with perhaps one backup).
It just doesn’t seem like much is being gained by individualizing perception in this case. Perceptive or not perceptive is an uninteresting character trait for purposes of roleplaying; it is pretty much just system mechanics. Does this same line of reasoning apply to hearing noise and searching for secret doors? Well, there is one major difference that I can think of. Listening at a door and searching are both proactive. In contrast, perception (or surprise) checks are generally passive, in the sense that they happen to the PCs rather than being initiated by them, and are also part of combat, which benefits more than other parts of the game from mechanical streamlining (because there are so many die rolls involved). If you base the search chance on class (like some editions do where default is 1 in 6 but elves have 2 in 6 change) then you minimize (but do not completely avoid) the skill tax problem.
I will close with one last observation. In the 90s, as has been commonly observed, tabletop RPGs moved away from problem solving and exploration and towards character development (influenced by the White Wold games and products that started to focus on character options). Thus, the logic of the individual character was privileged over the group (and game) experience. I can’t help but think that small mechanical changes like the perception skill are but manifestations of this larger trend, and are closely tied to the idea of RPGs as wish fulfillment fantasies. That, however, is probably a larger topic.