I have access to an unpublished PhD thesis in which a standardised personality test was administered to fluent German/English bilinguals. Even though the candidates scored equally well in the language proficiency tests, the personality test results proved to be a strong predictor as to which language the test was being taken in, as well as a strong predictor of whether the candidate's second language was compound or coordinate.<p>The evidence is stacking up that some moderate form of the Sapir-Worf hypothesis is true, despite what Chomsky and Pinker say.<p>The implications for programming langauges seem clear, but unsurprising. Computer languages have often been considered equally powerfully, but not equally expressive. Just as groups of polyglots shift the language used in discussions according to the topic, so the task at hand often suggests one programming language rather than another.<p>I'm still not sure why linguists (as opposed to polyglots) deny this.