Short term memory and peripheral vision at junctions

Motorcyclists are at extremely high risk of death in crashes where another vehicle pulls out into their path at a junction. Such crashes have often been described as the result of “look but fail to see” errors. However, recent research has shown that such errors occur even after a driver has looked directly at an oncoming motorcyclist. An alternative explanation for some of these crashes is the “saw but forgot” error. The idea that drivers might forget vehicles they have looked at only moments earlier is surprising, but it matches recent cognitive research that highlights limitations in visual short term memory. We present a cognitive model that highlights the limitations of drivers’ vision and memory at junctions. In two laboratory experiments we explore 68 young drivers’ abilities to make safe decisions at junctions and their memory for oncoming vehicles. In each experiment the driver is required to make head movements to view static images to the right, the left, and straight ahead. Views are projected over 180 degrees and the driver's basic task it to respond when they feel the junction is safe for them to cross. On occasional trials drivers are given a surprise memory test. Drivers find this task remarkably hard, frequently missing safe crossing opportunities and choosing to cross when it is not safe. Notably, drivers are much more likely to choose to a cross in front of an oncoming vehicle once they have looked away from it, suggesting that in this task failing to remember an oncoming vehicle is a more frequent cause of errors than failing to perceive it. This conclusion is supported by the memory tests which revealed that in even moderately busy traffic conditions the junction crossing task can exceed the capacity of visual short term memory. In the first experiment we separately explore the influence of peripheral vision, finding that drivers’ decision making and memory in this task is poor with or without peripheral information being present. In the second experiment we explore an initial intervention designed to enhance phonological memory but do not find that it improves memory for oncoming motorcyclists. In the second experiment we additionally separate out the influence of load (number of vehicles present) and delay (how long ago the vehicles were seen). This separation shows clear effects of load, but not delay on memory. We interpret these as evidence for both proactive and retroactive interference in visual short term memory. The current research shows that even a simple junction crossing task can exceed the capacity of visual short term memory. We suggest that this may be an important cause of junction crashes