It was my understanding that the author was using an imaginary object to make it easier to explain the concepts, as in the article he says:
It is possible to represent each execution context conceptually as an object with 3 properties
and goes on to talk about the activation / variable object.