In this presentation I discuss and demonstrate the advantages of using large natural language corpora for investigating regional linguistic variation and change. In the first part of this presentation, I describe an analysis of regional grammatical variation in a 36 million word corpus of American letters to the editor collected between 2000-2013. By mapping hundreds of grammatical forms across this corpus and by subjecting these maps to quantitative analysis, I identify common patterns of regional grammatical variation in written Standard American English. In the second part of this presentation, I describe an analysis of lexical spread in American English based on an 8.9 billion word corpus of geo-coded Twitter data collected in 2014. I first introduce a method for identifying newly emerging words in large time-stamped corpora. I then use this method to identify dozens of emerging words in the Twitter corpus. Finally, I identify common patterns of lexical spread in Modern American English by mapping the spread of these emerging words across the United States. To conclude this presentation, I argue that adopting a true corpus-based approach to dialectology, as opposed to using traditional methods for data collection such as surveys and interviews, allows for new and important research questions in geolinguistic to be pursued for the first time.