A supporting piece of research is the smoke-filled room study, by Latane and Darley. They found that when participants were alone in a room that began to fill with smoke 75% had gone for help by six minutes, 50% by four minutes, however, when there were three participants together, this dropped to 38%. This may have been due to diffusion of responsibility, whereby the participants did not assume personal responsibility (stage 3).
A drawback of this model is that it doesn’t take into account emotional factors, such as fear or anxiety, because it suggests a rational process of decision-making. Therefore, it ignores other factors which may affect whether or not someone helps, including gut reactions.
A second explanation of bystander behaviour is the Arousal: Cost-Reward Model (Piliavin). This suggests that emergencies are physiologically arousing (which is unpleasant), and if this is attributed to the person in distress, we seek to reduce it by helping them. The higher the arousal, for example if the bystander can empathise with the victim or is physically close to them, the more likely they are to help. The second part of the model suggests that before people offer help, they carry out a cost-reward evaluation, weighing up the pros and cons of helping versus not helping. If the cost of helping is high (e.g. risk of injury) we are less likely to help, unless the rewards are particularly high. If the cost of not helping is high, or the cost of helping is very low you are more likely to help.
One study which supports this theory, is Clarke’s cost of helping study. They found that students were most likely to agree to read to a blind student if they only had to travel a short distance (rather than a long one) and the student was preparing for an important exam. In this case, the cost of helping is low and the cost of not helping is high, as Piliavin’s model suggests.
Evidence supporting this model comes from Manstead et al, who found that people are aroused by the distress of others. Therefore the arousal part of this model is validated.
