For a real world example of a visual language vs. text, consider http://code.google.com/p/visual-lambda/.  Although this tool allows "intuitive manipulation" of lambda calculus, it's a poor substitute for reading plain text. 

Consider the operation 3 X 2 in lambda calculus text notation:

λ(xyz:x(yz)) 3 2

vs.

 

Pretty interesting, and maybe it could be used to manipulate expressions to understand reduction and so on, but useless for reading code and writing it on a large scale. 
I give this example for lambda calculus because it's probably the simplest possible.  The entire language is very simply expressed and (I think - I might be wrong here) all of the possible operations are included in this example.

EDIT:  To be fair, the visual lambda example is showing more than the text ... the text doesn't show all the reductions happening ... it's static, while the visual example is dynamic and reduces the expression.  Nevertheless, if you list out the reductions for the text, you'd still have an easier time in the end.