Wintertime cold snaps and summertime heat waves increase energy demand and draw heavily on emergency resources of state and local governments. Adequate planning for these events requires improved predictions on timescales beyond the short range where numerical models perform well. Comprehensive probabilistic tools relating temperature extremes to weather/climate conditions on multiple time scales from the extended range to seasonal-scales and longer would be highly useful to plan for these types of events. We have quantified heat waves and cold snaps for different regions of the U.S. over a 60-year period and used a probabilistic approach to relate these historic events to precursor weather patterns. Using atmospheric data from NCEP Reanalysis, we identified synoptic circulation patterns (predictors) that precede extreme cold/heat events at various lead times in the range of 0-30 days. By studying the evolution of predictor patterns, we find subtle but important differences in the atmospheric states that lead to an extreme temperature event versus those that are not followed by such an event. In some cases, low-frequency climate forcing appears to modulate the likelihood of an extreme temperature event developing, which may provide a link between seasonal and subseasonal scales. To address long-term planning, we also examine climate model simulations for their ability to realistically simulate synoptic causes of heat waves and cold snaps. We then look at spatially resolved changes in the evolution of temperature extremes under different climate change scenarios in the most realistic models.