We find a puzzling correlation in the data on domestic violence and children's outcomes in India. Using the 2005-2006 National Family and Health Survey, we see that girls in families experiencing spousal violence are less worse off than boys when only fathers report a son preference while the gender bias reverses when only mothers report having a son preference. To shed light on the puzzle in the data, we present a noncooperative theoretical framework based in economic theories of domestic violence, whereby differing parental son preference and bargaining over investments in girl and boy children potentially explains the observed relationship.