Abstractions are for pockets; they are miniatures of the world that we can carry around with us, that we can take out at our leisure and examine, and that we can tinker with. We can poke them and probe them and rearrange their parts. In essence, they are pocket toys.

Abstractions are pocket models of the world, and the scientific abstractions are a special class of these pocket models. For the scientist, abstractions must be useful models of the real world – the scientist would like to ensure that what he learns from tinkering with an abstraction will lead him to understand parts of the real world that he has not directly put his hands on. Scientific abstractions must have tiny portals that are windows into the unknown. For this to hold true, for a scientist's abstractions to enable him to see beyond his reference texts, he must be able to generalize from observations and experiments on the abstraction to observations and experiments in the real world. Thus, certain relations must exist between the abstraction and the real world phenomena and certain internal relations must also exist between the parts of the abstraction.