I have a theory that my car is white. The facts are that my car was white when I was buying it. Yesterday it was white, and every day earlier I saw it being white. Now the prediction of my theory is that tomorrow my car will still be white, because I have a law in my theory that my car is white always. This theory is not reality, it is just a theory. Because it might happen that the car will get painted to some other colour. In that case my theory would fail. The law would be broken by reality.
My theory is that the words: theory, knowledge, memory, word, concept have one root of meaning. Every word in a language is some kind of theory with its predictive power. We use language constructs to model the reality only eventually to be able to make predictions. In situations when we can descent to greater details of the model, more precise description is possible; for example, science theories. In other cases when the model is difficult to break into composites, we invent a concept, label it with a word, and use that word in a context to pass the information about future events; for example, knowledge or thoughts. I think that my car is white. Here I pass the model of my perception of the car. You would be surprised to see me not surprised seeing that my car is not white, because I told you that I think it is white, unless I lied. Lie is another concept quite different from think. In your model of me you may assume that I may sometimes lie. So your prediction of future events consists of many theories and concepts put together. The scientific method is built in our language, which appears to be the only way to describe the reality, via models and theories even if the models and theories do not look like models and theories in their usual meaning.