It also helps if exceptions are made in similar ways. For example, in 3e there were a number of combat maneuvers a character could use against an opponent: trip, disarm, bull rush, sunder, feint, grapple, and so on. There were also a number of feats you could take to improve these, with either Combat Expertise or Power Attack as prerequisites. Most of these feats were similar in that they removed the AoO for trying one of these maneuvers and also providing a +4 bonus. Some also provided some other benefit – e.g. Improved Trip letting you get a free attack when knocking an opponent down, or Improved Disarm not risking getting disarmed back. But these extra benefits were different for each feat, making it kind of hard to remember what works in what way. And then there was Improved Feint, which instead made feinting a move action instead of a standard action – working in a completely different way.The key problem with exception based design is that players must be familiar with all the rules there's an exception to. The simpler the system, the easier this is to achieve, but the fewer exceptions which can be made, which often leads to adding additional rules to compensate, which creates even more rules to make exceptions to.
That means that each of these feats require a separate info "slot" in the brain. But if they all worked the same way (+4 and remove AoO), and maybe had a separate upgrade feat that let them do something else, that would be less cognitive load.
To some degree this can be accomplished by using keywords. Keywords generally increase the up-front cognitive load, but reduce it over time. For example, there are a number of 5e spells that specifically make noise that can be heard up to 300 feet away. In a system with more keywords and less natural language, those would have a keyword like "noisy" so you wouldn't have to concern yourself with the noisiness of thunderwave or knock separately – you just know that a noisy spell can be heard 300 ft away.