"I speak French to my ambassadors, English to my accountant, Italian to my mistress, Latin to my God, and German to my horse."
Frederick the Great
True multilingual people have exquisite control of what language comes out of their mouth at any given time. (In contrast, I always mishmash my languages, putting German prepositions into my French.) At a cognitive level, this skill involves hearing and understanding the language and formulating a reply, while suppressing the other languages, as if a "language switch" is at work. But in MRI images of these people during writing or speech, the brain activity patterns are very similar no matter what language is being used. This is probably because most of the brain concerned with meanings embedded in the language is going to be independent of the language used.
A recent paper in Science used a trick to try to locate brain regions which were directly related to the choice of language at the level of words and meanings. The authors had multilingual subjects read word pairs, in which the paired words either showed a close relationship (i.e. trout-salmon) or were not closely related (trout-horse). By varying whether the words within the pair were obtained from the same language or different languages, they sought to specifically trigger the brain region that coupled language to meaning. (There's no telling what havoc recent American english adoptees such as 'angst' or 'samizdat' would wreak in this test!)The testees were wired up to PET scans or functional MRI to measure their brain activity during these tests.
Now this test is still not so simple, because differences in shapes and lengths of the words (German words were on average 7% longer than the English equivalents) in different languages will affect brain areas without being specifically concerned with the meaning of the words. Despite these difficulties, the scientists were able to see two new effects with this test: an area in the left temporal lobe was activated differently depending on the relationship between the word pairs, without being affected by the language; and activation in an area called the left caudate was reduced in same-language pairs compared to different-language pairs.
The left caudate is a very interesting candidate location for an internal "language switch" because of earlier data from patients with damage near this brain area. These people can understand languages, but spontaneously switch between languages in their own speech and writing.
I would be interested to see the social aspects of this language switch, as hinted at by the quote from Frederick the Great. There must be some sort of recognition for what language is best for a given audience. My wife and have I frequently noticed that peoples' speaking styles differ depending on the language. A person can be fairly formal and courtly in French, for example, and quite casual in English. I am not in full command of any language besides English, but it seems that a sort of style or swing-- a cultural expectation-- attaches itself to languages.
UPDATE: My wife pointed out that languages don't just have words but also grammar- which I again muddle, speaking French with German word-order. A little discussion of word-order differences in Basque-Spanish bilinguals is available here although I couldn't find a publication.