That's a huge debate whether seeing is really believing .. since first most people need to learn to see and at the same time others try to obstruct a clear view. Marketing is good example.
Back to topic: the article deals with an abstraction: Signal to Noise Ratio (SNR) which is compared between different sensors. In addition, to make comparisons possible, normalization is used (involving some Maths). How do you think this can be shown in pictures? The usual "100% crop" won't help here.
Secondly, when only looking at numbers, one thing is missing to me: At which level (difference in points or %) will the difference become visible to human eye under certain conditions? When testing sensors there is this remark that differences below 5 points are not visible to human eye (I suspect it's meant as "trained human eye" and consumers might only notice something at 10 points and more..).
I think a good gauge is to compare the results (same size) in final prints, that's what a photo is for. After all, whatever theory that is being used for the comparison has to be put into practical form for everyone to see. Isn't it? :think: