This paper examines detailed records from the civil conflict in Guatemala between 1977 and 1986. It reveals a number of novel patterns which support the use of complex systems methods for understanding civil violence. It finds a surprising, non-linear relationship between ethnic mix and killing; thereby inviting analysis based on group dynamics. It shows the temporal texture of the conflict to be far from smooth, with a power spectrum that closely resembles that of other, better understood, complex systems. The distribution of incident sizes within the data seems to fall into two distinct sets, one of which, corresponding to "regular" conflict, is Zipf distributed, the other of which includes acts of genocide and is distributed differently. This difference may indicate that that agents of the state were proceeding under different types of orders. These results provide an empirical benchmark for the modeling of civil violence and may have implications for conflict prevention, peace keeping, and the post-conflict analysis of command structures.