This paper uses novel data on English judges to explore whether the rule of law is above the influence of social interactions. Our research design has two key features. First, we focus on directed decisions affecting the well-being of another judge, and on the effect of interactions with that affected judge. Second, we focus on panels hearing unrelated cases in the Court of Appeal which are instances of social groups which form and dissolve with high frequency. These rapid group transitions are central to our identification strategy since they enable us to hold unobservables fixed by comparing decisions taken shortly before a social interaction with decisions taken shortly after. Together, these features circumvent the often cited problems (Manski, 1993) inherent to the identification of social interaction effects. Our results suggest that a judge who knows he will work with the individual whose case he is considering is substantially more likely to affirm than a judge who knows he will not, and hence that social interactions can indeed influence judicial decisions.