If this were the case, one would expect the hidden seasonal variable to exert its influence not only during the twilight zone test periods but also at other times of day.
In Quasi-Experiment 2 we investigated this possibility by examining seasonal variations in the pattern of accidents that occur in daylight control periods as compared with the twilight zone test periods.
We reasoned that most seasonal variables, other than light condition, should affect accident patterns in the daylight control period as well as the twilight zone.
For example, if severe weather were responsible for increased fatal collisions in the winter, one would expect parallel increases of fatality rates in both the daylight control and the twilight zone test periods.
Quasi-Experiment 1 used yearly proportions of daylight and darkness during the twilight zone test periods as a chance prediction of the likelihood of fatal accidents that would be coded as occurring in darkness.
The results of Quasi-Experiment 1 showed that the incidence of road fatalities is disproportionately higher during the darker than the lighter phases of the twilight zone test periods.
We emphasize that with respect to alcohol, the present results reflect only the situation encountered in the twilight zone test periods; higher levels of alcohol involvement are likely in periods closer to midnight, and lower levels in periods closer to noon.
However, there was virtually no concurrent variation in the monthly distribution of other fatal accidents in the twilight zone test periods, indicating that reduced natural illumination is not an important contributor to this class of events.
That possibility seems unlikely, however, because the hidden variable would have to be (a) systematically confounded with illumination from the sky during the twilight zone test periods and (b) selectively related to fatal pedestrian/pedalcycle accidents.
The twilight zones comprise two times of day, morning and evening, during which variables such as traffic density, driver age-mix, fatigue, and alcohol consumption should be relatively stable year round.