Recall that one of the parameters in the design procedure [for highway crest curves] is the height of the obstacle to be seen by the driver in time. Originally (already in 1940) American engineering standards set the obstacle height at 4". Those who wrote this standard did not have any particular obstacle in mind (although, rumour has it that some refer to it as the ‘dead dog’ criterion). …the 4" was selected not because lower obstacles are not a threat to safety but because the selection of a higher obstacle would not save much in construction cost. Since, at that time, nobody knew how many crashes are due to obstacles on the road, what kinds of obstacles these are, and what fraction of crashes would not have occurred had the crest been flatter, the standards committee did what was sensible. They made a decision on the basis of what was known, namely the cost of construction.
For two decades everybody was designing roads using exacting calculations to make 4"-high obstacles visible in time to stop. Then, around 1961, it became apparent that in the newer model cars the average driver's eye was much lower than a decade or two earlier. Thus, drivers of newer cars could not really see 4" objects at the prescribed stopping sight distance... The solution to the predicament was not difficult. Since the 4" obstacle neither was some real object nor has it been selected on the basis of some factual relationship to safety, the Committee on Planning and Design Policies had no compunction noting that "the loss in sight distance resulting from lower eye height could be offset . . . by assuming an object higher than four inches . . . ” Indeed in the 1965 AASHO Blue Book, 6" obstacles became the standard of design.
At the time the standard for the design of crest curves came into being, little was known about safety. Today we know that only 0.07 percent of reported crashes involve objects less than 6" high. We also know that, till today, no link has been found between the risk of collisions with small fixed objects on crest curves and the available sight distance. On the contrary, “Crash rates on rural two-lane highways with limited stopping sight distance (at crest curves) are similar to the crash rates on all rural highways.” Thus, the assumption invoked at the dawn of highway design history which allowed the formulation of a design procedure based on the avoidance of dead dogs in the middle of the road seems to have little to do with real road safety. Still, till today, the same standard stands, the same exacting but illusory constructs are used in the design of crest curves. Only the size of the dog and of other parameters is changing.