This paper demonstrates the potential of sequence analysis (SA) for conflict studies by applying this methodology to a new dataset on violence by US counterinsurgency forces in Vietnam. We discover patterns and repertoires of violence in this data that conventional regression-based techniques of would not normally detect. As these methods do not pay close attention to the individual trajectories of units of analysis over time, they analyze them in a relatively context-free fashion; yet it is within these trajectories - the sequences of states that the units of analysis undergo throughout - that we find detailed information about conflict processes: what kind of events precede others, what consequences certain events have, which chains of events are typical of particular conflicts. SA can analyze these trajectories - encoded as sequences of categorical states - individually, collectively, comparatively or for subsequences. It is a versatile technique for discovering, describing and explaining patterns and variation within and across sequences and hence a universal methodology for discovering patterns and trends, antecedent conditions and consequences of certain events or event types within the trajectories of units of analysis. SA can thus analyze "repertoires of violence" - a notion frequently invoked but still loosely conceptualized. Such repertoires can be defined as the set of types and relative proportions of violence against non-combatants that armed groups regularly employ. For our case, we describe these repertoires and discover that they relied on clearly patterned rules based on specific interactions preceding the actual violence. These rules are easily identified by SA but difficult to discover using regression-based techniques. Furthermore, SA is computer-aided and therefore systematic and replicable. We use the free statistical environment R and its TraMineR package for the entire range of our SA, from detecting and visualizing sequences to discovering patterns and analyzing these statistically and comparatively.
