For high-rise building evacuations, travel on stairs accounts for a large portion of the total egress time. Previous research has typically focused on the relationship between speed and density, with the results being relatively poor for predicting how long an individual requires to descend the stairs.

In this dissertation, methods for measuring fundamental variables for determining speed and density are presented. The inverse of speed (normalized time) is shown to be more conservative for prediction methods. Furthermore, the interactions between individuals and flow units and the characteristics of individuals provide a deeper understanding of the flow dynamics and lead to more accurate predictions than relying on density alone.

Rather than all of the occupants behaving in a uniform manner, flow units form where all of the occupants in them descend at the same rate. The first persons in flow units were found to be engaging in five different types of behavior that set the pace for their followers. There are also flow states identified that previous researchers have not identified.

How occupants interacted with one another was also found to vary based on individual characteristics. The behavior of individuals was not random, but appeared to be based on variables like gender and exit lane. These characteristics, in turn, were found to influence their descent rate.

These findings are then applied to a blind data set and the actual observations are accurately predicted. This allowed equations to be presented that are representative of the flow dynamics.