I think the problem overloads the term knowledge. First, there's knowledge about the external world, the kind which enables you to predict the conseqeuences of

I think the problem overloads the term knowledge. First, there's knowledge about the external world, the kind which enables you to predict the conseqeuences of your actions and to manipulate the world to achieve desired results. Concretely, Mary will be able to use a, say, hex editor on a BW terminal to "paint" a picutre which a "normal" will recognize as a, say, red poppy flower. She manipulated another person into having the experience of seeing a RED poppy flower. Second, there's knowledge about your internal world, how you experience things. If Mary is shown the poppy flower picture on a color monitor, she will gain new, experiential, knowledge about how her own body reacts to that particular stimulus. It's a knowledge, but it's not a knowledge that gives her abilities to manipulate the external world in new ways. The premise of the experiment doesn't account for the fact that each individual is a sum of their past experiences, encoded physically in their brains in some way. One person upon hearing "red" will think about tomatoes, another about blood, yet another about poppy flowers. (I grew up at a place where they were common and we as kids even ate the seeds :)) Thus the premise is false, the fallcy is in the formulation "what goes on when we see ripe tomatoes"; emphasis on we. There is no collective "we" in this case; she doesn't know all physical information shee needs to know in order to manipulate the external world (e.g., knowing whether to draw a poppy flower or a tomato in order to evoke a particular emotional response from a particular individual).