The overall goal of the SHRP 2 is to prevent or reduce the severity of crashes by understanding driver behavior. Integrating the tasks outlined in this research project will result in advanced knowledge of driver behavior, enabling those in the field to enhance both current in-vehicle telematics and roadway infrastructure. Ultimately, this will lead to reduced severity and numbers of vehicular crashes. We begin by examining the interaction between data captured from the roadway environment, driver, and the vehicle. Each influences the other, with the output informing exposure and risk, as well as crash surrogates. Traditional regression techniques are often used to predict crash likelihood. To fully examine the complex dimensions of crash risk, however, requires advanced statistical techniques to discover factors that are similar to crashes, to examine crash surrogates using appropriate exposure measures, and to describe crashes using dynamic modeling techniques. A complex array of factors contributes to crashes, which this team will address with a systems-based approach. A clear set of research questions and objectives will guide the data collection process and provide the framework for the analytical method. This analytic framework may include exposure-based risk estimates within each naturalistic data collection option (i.e., in-vehicle, or site-based). A well-defined set of explanatory and dependent measures will structure the research approach to assess risk. There is a vital need not only to clearly define driver performance measures for naturalistic driving, but also to assist practitioners in understanding how complex data can be used to help formulate sound, evidenced-based policies. Clear research questions and an analysis plan that captures the core issues of driving safety utilizing similarity discovery, spatial and temporal analyses, and dynamic modeling will be developed.