In general, the unsafe conditions that are likely to produce a motor vehicle crash reside not at the mean of a given distribution (in other words, under typical conditions), but rather in the tails of the distribution. For example, an unusually slow response to a traffic obstacle, rather than an average response, may result in a collision. Although that situation means that crashes are the exception and not the norm, it has implications for how safety-critical data are approached and handled. In this current paper, experimental data collected in a driving simulator are used to demonstrate how an analysis of the average glance durations to an in-vehicle display might lead to different conclusions about safety compared with an alternative analysis of the tail end of the distribution. In addition, a model of crash risk based on the distribution of in-vehicle glances is described, as well as several characteristics of the traffic environment.