Working memory, also known as short-term memory, is the part of the human memory that is active in a given moment. The following model of how it works was apparently put forward by Alan Baddeley and Graham Hitch in 1974.
It consists of four systems: a central executive, which is like the “boss” system that manages and coordinates all the others, and three subordinates: the phonological loop, which stores phonological content (basically, words), the visuo-spatial sketchpad, which stores visuo-spatial data, and the episodic buffer, which synthesises information from the previous two subordinate systems and long-term memory into, as the name suggests, basically a play-by-play of what’s been happening lately. The phonological loop is also sometimes called echoic memory, because basically it “loops” an “echo” of about 3–4 seconds’ worth of phonological information. (This is why sometimes you might not understand something on first listen, but then it “clicks” a few seconds later after you’ve replayed it in your mind.)
The reason Baddeley & Hitch identified the phonological loop and visuo-spatial sketchpad as two separate systems is that studies have shown that when someone is multitasking on two separate tasks that involve those two separate systems separately (i.e. a verbal task and a physical labour task), their performance is almost as good as if they were focusing solely on one task. However, if trying to multitask on two things that involve the same system at once (e.g. listening to a podcast and reading a book at once; or mentally mapping out a route while trying to do chores), performance is heavily compromised.
The episodic buffer was not identified as a separate subsystem until 25 years after the rest of the model was described, but it was added because it had been noticed that amnesiacs (lacking long-term memory) were often much better at describing episodic memories (e.g. the events of a movie they were watching) than the previous model of short-term memory would have enabled them to do.