My understanding is that originally affordances meant the interactions that are physically possible owing to the form of an object. When we say a doorknob two inches in diameter affords single-hand grasping we simply mean it possible to grasp it. If it were two feet in diameter, it would not afford one-hand grasping. From our experiences, we humans learn to recognize affordance by appearance, so this can be used to guide interactions.
The original concept of affordances falls apart when you get to software user interfaces because there are no physical interactions and anything is in principle possible. A user can click anywhere. It’s just a question of if and how something will respond. In the essay you linked to, Norman attempted to clarify and expand the concept of affordances to include “perceived affordances” in addition to the real affordances that are only possible with physical interactions.
Perceive affordances, according to Norman, are “cultural constraints” and “learned conventions.” Underlined colored text “affords” clicking because users have learned that’s what indicates a link. In other words, having perceived affordances is simply being consistent with what everyone else does (and therefore what users have learned). It merely means following tradition and well-known standards. Can something afford toggling? If it looks like how toggling is traditionally distinguished from non-toggling (e.g., a checkbox), then yes. Likewise for selectable.
The concept of perceived affordance lacks rigor. How conventional does a convention have to be to qualify as an affordance? The usefulness of the concept is questionable. Why use the term “affordance” when we already have the term “consistency”? I sometimes find “affordance” useful to refer semi-formally to a metaphor, convention, or standard so ingrained in user habits as they respond to it automatically, almost compulsively. But arguments over precisely what is or isn’t an affordance are irresolvable.
The vagueness of the affordance concept can also be counterproductive. I get the impression the concept has led some UX designers to believe there are certain elemental perceptual properties that inherently communicate interactions to the user, such as gradient shading affords clicking to select, or a list of objects (on a mobile device) affords swiping to scroll. However, if we recognize affordances as synonymous with external consistency, we realize there are no elemental perceptual properties. Instead there is the holistic image and its similarity on multiple dimensions with previous experience. Gradient shading only works reliably on a smallish long rectangularish shape with centered text label, for example.
I would not object to “affordance” being deleted from our professional vocabulary like “user friendly” has been.