Listening matters because acquisition comes from input, and input helps only when it is mostly understood. If speech is too difficult, attention goes to decoding sounds, so little is left for picking up new words.
Picture a learner hearing the thief crept through the window. Because thief and window are familiar, context lets her infer crept, and repeated hearings gradually fix the meaning.
How one learner's listening session builds vocabulary
The thief crept through the window is where the case starts. The learner already knows thief, window, and through, so she has spare attention while listening. She uses it to guess that crept means moving quietly, because the story context makes that the only fitting meaning.
A few days later the same narrator says the cat crept onto the sofa so nobody would see it. Because the sentence structure and the other words feel easy, she links the new hearing to her earlier guess. The meaning sharpens: quiet, careful movement to avoid being noticed.
Each replay of the story adds one small stretch, like a slightly harder word or a faster passage, while most of the language stays familiar. So each session costs little effort but builds a little more, and crept moves from guessed to reliably known over repeated hearings.
The mechanism works because guessing a word from context needs the surrounding words to be mostly known. If the whole sentence had been full of unknown words, there would be nothing left to anchor the guess, so crept would just stay a blur of sound.
What the evidence actually covers
One limit deserves a plain statement. The familiar findings about guessing new words from context, and about how much of the language needs to be known before a guess works, come mostly from studies of reading.
So the picture of crept being inferred through easy surroundings is a plausible extension to listening. The studies behind these findings mostly tested reading, so listening-specific figures, like exactly how familiar speech must be before a learner can guess a new word from context, are not yet established.
A second limit touches the mechanism itself. The idea that hard input eats attention, leaving little capacity for acquisition, is a reasonable model rather than a measured listening result. The same caution applies to the opening idea of this explanation, that mostly understood input frees attention for picking up new words. For you as a learner, treat that picture as well-motivated, but not as a precisely tested figure for how much overload blocks learning during listening.