This paper compares the rates of use of phonological variation across speech and, through variant spellings, Twitter data. Speech data came from New York sports personality Mike Francesa and one of his fans parodying him. The fan's tweets, which include a large number of variant spellings, were also analyzed. Two patterns emerged. For the most salient variables, th-stopping and r deletion, the fan overshot Francesa's use in speech and used them at that same higher rate in his tweets. However, for less-salient variables the fan used them at the same rate in speech as Francesa and used them at a far lower rate in their tweets. This suggests that the rate of use of a variable on Twitter can be used to investigate salience.