I've grown tired of reading study after study and the many different findings that all seem to contradict one another or are just totally ridiculous. So I decide to call the National Safety Council and ask them "What is the most conclusive data you've got?" ... Unfortunately the person I really need to speak to was already gone for the day but the person I DID speak to (who admitted they were knowledgeable but not an expert on the topic) recommended I start by looking at a recent Virginia Tech study.
After starting to read it I realized that I've already been through it but I do agree with some of the opinions they offer in this part:
The Disconnect Between Naturalistic and Simulator Research
It is important to keep in mind that a driving simulator is not actual driving. Driving simulators engage participants in tracking tasks in a laboratory. As such, researchers that conduct simulator studies must be cautious when suggesting that conclusions based on simulator studies are applicable to actual driving. With the introduction of naturalistic driving studies that record drivers (through continuous video and kinematic sensors) in actual driving situations, we now have a scientific method to study driver behavior in real‐world driving conditions in the presence of real‐world daily pressures. As such, if the point of transportation safety research is to understand driver behavior in the real‐world (e.g., increase crash risk due to cell phone use), and when conflicting findings occur between naturalistic studies and simulator studies, findings from the real‐world, and not the simulator‐world, must be considered the gold standard.
It is also critical to note that some results of recent naturalistic driving studies, including those highlighted here as well as others (e.g., Sayer, Devonshire and Flanagan, 2007) are at odds with results obtained from simulator studies. Future research is necessary to explore the reasons why simulator studies sometimes do not reflect studies conducted in actual driving conditions (i.e., the full context of the driving environment). It may be, as Sayer, Devonshire and Flanagan (2007) note, that controlled investigations cannot account for driver choice behavior and risk perception as it actually occurs in real‐world driving. If this assessment is accurate, the generalizability of simulator findings, at least in some cases, may be greatly limited outside of the simulated environment.