•Naturalistic driving study data are used to assess the relationship between mean speed, speed variance, and crash risk.•Speed selection is impacted by posted limits, though the changes in speed are markedly less than the difference in limits.•Crash risk is primarily influenced by variability in speeds, both directly and indirectly through other factors.•Individual drivers tend to exhibit correlation in both speed selection behavior and crash/near-crash risk. This study investigates how speed limits affect driver speed selection, as well as the related crash risk, while controlling for various confounding factors such as traffic volumes and roadway geometry. Data from a naturalistic driving study are used to examine how driver speed selection varies among freeways with different posted speed limits, as well as how the likelihood of crash/near-crash events change with respect to mean speed and standard deviation. Regression models are estimated to assess three measures of interest: the average speed of vehicles during the time preceding crash/near-crash and baseline (i.e., normal) driving events; the variation in travel speeds leading up to each event as quantified by the standard deviation in speeds over this period; and the probability of a specific event resulting in a crash/near-crash based on speed selection and other factors. Speeds were relatively stable across levels-of-service A and B, within a range of 1.5mph on average. Speeds were marginally lower (3.3mph) on freeways posted at 65mph versus 70mph. In comparison, speeds were approximately 10.2 to 13.4mph lower on facilities posted at 55mph or 60mph. Speeds were shown to be 2.5mph lower in rainy weather and 11mph lower under snow or sleet. Significant correlation was observed with respect to speed selection behavior among the same individuals. Mean speeds are shown to increase with speed limits. However, these increases are less pronounced at higher speed limits. Drivers tend to reduce their travel speeds in presence of junctions and work zones, under adverse weather conditions, and particularly under heavy congestion. Crash risk increased with the standard deviation in speed, as well as on vertical curves and ramp junctions, and among the youngest and oldest age groups of drivers.