I'd rather say "how well internalized shapes are".

I feel like drawing pictures today, so I thought I'd explain why I think of internalized patterns as reading, if you were able to read at one time.

To me, when somebody says that they "read" something, it means that they are reading the game tree of possible moves.

Initially, the gametree may look something like this:



There are a number of nodes that could be possible solutions. From the root of the tree, it is unknown as to where the (green) solution is. You just don't know.

So you practice a lot of go problems, and you start to learn what paths in this tree "work". You start to be able to prune branches so that, later, when you look at the tree, you know that certain branches are futile. You prune branches like this:



After you've done this process really well, you are left with a pruned gametree, which is like this:



Yes, you don't have to traverse any nodes to reach the "final" solution, but you are still reading from the gametree. It's just that your gametree is nicely tidied up

_________________
be immersed

For tsumego (or similar things like cut/connect problems) I totally agree. You read something out, you remember the status, and next time you can stop reading when you reach the same situation. That's pure reading. Or, from a more practical point of view that doesn't depend on semantics: That's what you learn when you do tsumego. Both the reading through the tree, and the final shapes where you simply know the status.

For middle game reading I think it's a bit different. Of course, you still have to stop going down the tree at some point, and you have to decide whether it's a good or a bad line. But in addition to the obvious good or bad lines (e.g. a group died), I don't think you can ever arrive at a definite answer by reading alone. You can however see how the game turns out. That's more like a subconcious statistical approach. In the typical middle game situation, it's very likely you'll never get a definite feedback like in life and death situations. Not on the board, and also not in your read out variations. The only feedback you'll get is indirect, for example when you'll have to make life locally in gote. But recognizing that's bad is already something strategic. In the end, the only way to find that out on your own is to observe that you typically lose games where it happens to you, and win games where it happens to the opponent.

These things I definitely wouldn't call reading. Not only don't you scan a game tree, you never did. You didn't arrive at the result by reading, you arrived at the result by observing results. That's what I would call "positional judgement". Again, stepping back from the semantics discussion: Regardless of whether you like to call that reading, I don't think you can learn that skill from tsumego.

Note that you can aquire both skills from theory books and on your own. You can either believe the books that the L-shape is dead, or you can read it out yourself. And you can believe the books that it's bad to get shut in and live locally, or you can try it out in your own games and see how it works out. But you can't really "read it out". Two different skills, and two different methods to learn them

Of course in practice, in a game, you don't do any of this. You just stop reading at both the L-shape and at the group that is shut in, and label it as bad (or good) variation.

EDIT: And before someone complains: Yes, I do realize it's not always bad to get shut in and live. Recognizing these exceptions is another example of something that you can't do by just reading it out.