Abstract
Every syntactician is accustomed to experimenting with a given sentence. We modify it slightly, adding or removing a negation, changing the word order, replacing one word with another. Then we evaluate the result of the modification.
Every syntactician would like to be able to do the same with a language, to modify it slightly, to change one of its properties, to see what the result is.
Like all scientists, syntacticians have to accept that they can't do everything. But this can be overcome by studying a set of closely related languages, in such a way as to bring out syntactic properties that are so strongly correlated with each other that all the languages in question have either the presence or absence of these properties in common.
This is followed by a number of examples of such correlated properties of Romance languages, of great theoretical interest.
