Radio shows -- dramas, in particular -- employed both of these forms. Which type you got depended on the show. THE LONE RANGER and SUPERMAN used observer narration. The observer was particularly helpful to describe things that were going on that would be difficult to convey in dialogue or with audio cues -- in Superman's first radio adventure on Earth, for example, the narrator describes Superman as hovering in the air over a city in Indiana. The show used audio cues -- "Up, up, and awaaaaaayyyyy!!!" and the sound of rushing wind -- to convey Superman's flight, but conveying Superman hovering in space when he didn't have anybody to talk to there was a little more difficult. The narrator for THE LONE RANGER might describe a horse chase, or describe what various characters were getting up to offstage. Participant narration was common in the detective shows; DRAGNET is narrated by Jack Webb's Joe Friday, and YOURS TRULY, JOHNNY DOLLAR is narrated by the insurance investigator himself. (I think that's why we think of the first-person voiceover as such a part of the hard-boiled genre: it has less to do with film noir than with radio.)
So, where do comic book captions come into it? In the Golden Age, captions were used almost exclusively as third-person narration, to set the scene or describe complex actions. In the modern era, forget it. Captions are almost exclusively first-person participant narration. They've usurped the place of the observer. Another casualty of this change: the thought bubble, which you almost never see anymore. These didn't represent narrative injections; their role was more akin to asides. But that's another issue.
In the Golden Age, radio and motion pictures were the major forms of mass entertainment. Of the two, comics had more in common with radio: less running time, more limitations of format, more opportunities for freedom of imagination (no special effects budget). Narration was nearly ubiquitous in radio drama. Because it was *needed.* People couldn't see what was happening. If the Lone Ranger is chasing a bad guy on a horse, you can't see how close he's getting -- you need the narrator. If Johnny Dollar is in a fistfight, you have no idea how he's doing until he tells you, because all you hear are grunts and punches. As radio faded out and television came in, the narrator disappeared. He wasn't necessary. The Lone Ranger didn't need a narrator to tell you that he'd taken off his mask and disguised his face; you'd see the mask hanging over a tree branch, and Clayton Moore in a fake beard, and you knew perfectly well what was happening. Narrators went the way of the dodo.
Except for the detectives.
I don't know why this is, exactly, but I can make a bet on part of it, and that part is spelled: dun-dun-DUN-dun. Jack Webb started making DRAGNET on the radio, and kept making it on TV, and everybody on it knew the formula. And the narration stayed. First, because it helped sell you on what was going on, and second, because it helped sell you on Friday.
The change in the role of the caption boxes is a change in the generation of comic writers. The original comic writers were heavily influenced by the mass medium of their day: radio. There's no doubt that today's comic writers are overly influenced by the mass medium of their day: television. Hence decompression, and the six-issue storylines that Binder and Beck, or Kirby and Lee, would have handled in one issue. Many of today's comic book writers aren't writing comics any more. They're writing screenplays. And the shift away from the observer's captions are part of that.
But why are they still there? And why in first person? Why not do away with them entirely?
If I had to give a reason, I'd say: Frank Miller's BATMAN: YEAR ONE. If I had to pick a series that marked the changing role of the caption box, that'd be it. Miller's captions provided narration, in different voices: Bruce Wayne's handwritten captions, versus Jim Gordon's typewritten ones. It was astoundingly effective. And a lot of the use of participant narration in mainstream comics these days is probably due to that.
Why did Miller do it? Well, look at those characters. Think about what they have in common.
Batman and Jim Gordon are both detectives.
I think the changing role of the caption box relates to the changing use of narration in general in mass media, and reflects the change in generations of the writers, and the media they imprinted on. My guess is that Miller used the established convention of the detective-narrator for effect, to evoke a world that's a little hard-boiled, a little noir. I don't know if he even did it consciously, or if it's something that was culturally internalized, like the way every kid who plays at being a pirate unconsciously imitates Robert Newton in the 1950 production of TREASURE ISLAND. But I think that's what's going on, and I think Jack Webb is a big part of the survival of the trope. People who know more than I do, feel free to tell me I'm wrong!
For reference, here's observer narration in THE LONE RANGER, and here's participant narration in DRAGNET. Bonus: the LONE RANGER episode is *totally insane,* and Tonto's frequent long, drawn-out utterances of "MMmmmmm" make him sound a lot like Peter Boyle's portrayal of the monster in YOUNG FRANKENSTEIN.)