Rafael and Jose were awake every day before 4 AM to make their way across Boston to South Station, the city's main train station. There they would catch an outbound train to the wealthy suburbs to the west of the city, arriving in the village of Grafton before 6 AM. Then they would wait on the platform.When the next inbound train arrived, they would head back to Boston.

According to hundreds of people to whom I showed their photographs, Rafael and Jose looked handsome, intelligent, and friendly. In my personal interactions with them, I found them so. But other rail passengers at Grafton station were uncomfortable and seemed not to want them there.

How do I know this? Because I surveyed these other passengers and their responses were unambiguous. Why were they uncomfortable with these handsome, intelligent, and friendly-looking young men? Because they were speaking Spanish.

Rafael and Jose were at this train station because they were part of an experiment I conducted in the summer of 2012. My goal was to make people think that their neighborhood was experiencing demographic change due to immigration; that is, I wanted to change the social geography of their community. My research team placed pairs of people working for me – “confederates,” in the language of experimental research – at train stations to wait for the same train at the same time each morning during rush hour. These confederates were native Spanish speakers. A train time had been paired with another train time at the same station and, randomly – by the proverbial “flip of a coin” (actually by a pseudo-random number generator on my computer) – one time was assigned to treatment and one to control. The train time assigned to treatment was visited by the confederates; the other was not.