We present a series of experiments to understand the nature and extent of discrimination in urban Lima, Peru. The experiments exploit varying degrees of information on performance and personal characteristics as people sort into groups to test for statistical versus taste-based discrimination. This allows us to examine the nature of discrimination. Our sample is similar to the racial and socio-economic diversity of young adults in urban Lima. This allows us to look at the extent of discrimination. We use a unique method to measure race, along four racial dimensions common in Peru, and find that race is clearly observable. This gives us confidence that we can examine discrimination based on race. While behavior is not correlated with personal, socio-economic or racial characteristics, people do use personal characteristics to sort themselves into groups. Beauty is a robust predictor of being a desirable group member as is being a woman. Being unattractive or looking indigenous makes one less desirable and looking white increases one’s desirability. Interestingly, indigenous subjects are three times more likely to be classified as unattractive, suggesting that beauty might mask discrimination. We find that once information on performance is provided, almost all evidence of discrimination is eliminated, except in the most-preferred group. The evidence in these cases is consistent with taste-based, rather than statistical, discrimination.