Experimenters were collected from Columbia University General Studies class. It included four groups of four students, 2 female observers and 2 males-1 victim and 1 model. 3 victims were white, 1 was black, all aged 26-35 and dressed alike.
The experiments were conducted between April 15th and June 26th on 4,500 people travelling on the A and D trains of the 8th Avenue New York subway between 59th Street and 125th Street between 11am and 3pm. The journey lasted 7.5 minutes. 6-8 trials were conducted on any given day.
The victim either smelled of liquor or appeared sober and carried a cane. Models had 4 conditions-early/late critical area and early/late adjacent area. Two observers recorded DV. One noted race, sex, location of each person in critical area and of each person who helped and total number of people who helped. The other noted the same details but in adjacent area, latency of first sign of help and both noted comments made and tried to elicit comments.
Victim stood by a pole in critical area, after 70s he staggered forward and collapsed. Until someone helped he remained supine on the floor staring at the ceiling. If no help came by the time the train stopped the model helped. After each condition was compleye the team dissembarked seperately waiting for dispersion. They then travelled back the opposite way.
Cane victim received help 95% of the time compared to the drunk victim with 50%. Help was offered quicker to cane victim (median 5s compared to 109s for drunk victim). 60% of trials when help was given it was from 2 or more people, 90% of first helpers were male.
Slight tendency found to help same race especially in drunk condition. Diffusion of responsibility was not found. Individual who appears ill is more likely to get help than one who appears drunk. Men are more likely to help than women. When escape is not possible help is likely to be forthcoming. Bystanders conduct cost-benefit analysis before making a decision to help or not.